Apple Could Use ChatGPT or Claude to Power Siri

[SBlue1]

Yes please! Just make it an option.

 

[dwaite]

Why would they rely on OTHER companies for? It only proves that Apple cant make such things due to poor hardware and software. Besides, M3 Ultra is not even close to be "enough". Does M3 Ultra even close to Nvidia A100?

Depends entirely on the scale, task, and on software being optimized for the hardware. Both can do work the other falls over on.

A business can drop half a million on a single server loaded up with Nvidia cards, but they probably shouldn't do that just because someone heard Nvidia was 'great at AI' or the like.

 

[dwaite]

The cloud based ones are the ones being deployed on Apple servers with M3 Ultra or what I’ve also heard is M4 Ultra based servers because they run much cooler than M3s.

My information is quite dated at this point, so I will say my speculation is that they are limited in how much complexity they can shove into a package until they can adopt SoIC-mH. I expect the M3 Ultra to be replaced by the M5 Ultra.

 

[BNBMS]

Depends entirely on the scale, task, and on software being optimized for the hardware. Both can do work the other falls over on.

A business can drop half a million on a single server loaded up with Nvidia cards, but they probably shouldn't do that just because someone heard Nvidia was 'great at AI' or the like.

And yet, Apple is the one who is admitting that they cant compete in AI markets. Amazon, Microsoft, Tesla, Google, and more have their own servers/super computers and they even have their own chips. Acquiring companies will never work for AI markets as Apple dont know what to do with it while keep delaying Siri and Apple Intelligence features.

 

[NewOldStock]

Will copyrighted material ever be safe?
And will copyrighted material scraped from the internet be safe?
End of all copyrights as we know them? I know its been said and very sad intersection point in technology.
One giant data base if LLM for all and for free. Gui/UI only needed.
Sorry officer i did not steal that information.
Need more answers and not lawyers.

 

[MidwestMacGuy]

As some people hate to admit it, this is all because Apple didnt care about AI for a long time and yet they crippled AI teams thanks to Apple CFO. They cant even make their own AI server as Apple Silicon isn't powerful enough to replace Nvidia GPU. Since they cant even make Apple Silicon version of Mac Pro with powerful GPUs, they will not be able to research, develop, and train their own AI instead of relying on others just like they did with Intel for Mac.

More I see articles about Apple's AI team being upset and disappointed toward Apple, it only proves me right!

Someone needs to explain to me why you need powerful chips in a laptop to run AI. It seems most of the computing work is done in the cloud. I ask Claude a question, it searches it, database or the Internet and gives back an answer. Therefore, the computer is just displaying the results, and again, it seems to me all the processing work is done in the cloud. What am I missing here?

 

[AppliedMicro]

Someone needs to explain to me why you need powerful chips in a laptop to run AI. It seems most of the computing work is done in the cloud

AI tasks do not need to run in the cloud.
They can be run locally.

 

[dwaite]

Someone needs to explain to me why you need powerful chips in a laptop to run AI. It seems most of the computing work is done in the cloud. I ask Claude a question, it searches it, database or the Internet and gives back an answer. Therefore, the computer is just displaying the results, and again, it seems to me all the processing work is done in the cloud. What am I missing here?

Different architectures.

Claude runs a giant model in the cloud with access to internet stuffs and documents you specifically uploaded or prior chats. Everything you do first goes through that.

Siri AI runs a hybrid setup - a small model locally with access to everything spotlight indexed on the phone, and intents (actions) exposed by applications including internet search. If that small model can't determine the right answer, it packages up the question and the set of information it thought was necessary locally and sends that up to Apple servers for them to crunch using a (moderately giant) model.

 

[Siliconguy]

Someone needs to explain to me why you need powerful chips in a laptop to run AI. It seems most of the computing work is done in the cloud. I ask Claude a question, it searches it, database or the Internet and gives back an answer. Therefore, the computer is just displaying the results, and again, it seems to me all the processing work is done in the cloud. What am I missing here?

Not everyone wants to send their personal data to the cloud. So anything that can be done on your computer should be done on your computer.