macplusplus
About
- Username
- macplusplus
- Joined
- Visits
- 287
- Last Active
- Roles
- member
- Points
- 3,138
- Badges
- 1
- Posts
- 2,118
Reactions
-
Apple is lying about Apple Intelligence, John Gruber says -- and he's right
canukstorm said:macplusplus said:More personalized Siri... Easy to utter, extremely hard to conceive and implement. The biggest drawback of current LLMs is their lack of "context retention". Ask any of the most powerful LLMs they will list that among their limitations. Even in a single session they have difficulty on maintaining an established response pattern in repetitive tasks. A Siri that responds with a different personality everytime is intolerable. Maintaining the context is crucial even for an avatar-level "personality". Apple's refusal of a premature jump lnto the "AI smartphone" bandwagon has certainly serious technical reasons. Remember that Apple has already laid out a very solid foundation with the A18 chip designed specifically to run on-device LLMs, and before that, the Neural Engine. In that sense, the "lack of context" shared by Gruber and the author is even more amazing than that of LLMs.
Extraordinary claims require extraordinary evidence. Apple's lost trust. It's their job now, to earn it back. -
Apple is lying about Apple Intelligence, John Gruber says -- and he's right
More personalized Siri... Easy to utter, extremely hard to conceive and implement. The biggest drawback of current LLMs is their lack of "context retention". Ask any of the most powerful LLMs they will list that among their limitations. Even in a single session they have difficulty on maintaining an established response pattern in repetitive tasks. A Siri that responds with a different personality everytime is intolerable. Maintaining the context is crucial even for an avatar-level "personality". Apple's refusal of a premature jump lnto the "AI smartphone" bandwagon has certainly serious technical reasons. Remember that Apple has already laid out a very solid foundation with the A18 chip designed specifically to run on-device LLMs, and before that, the Neural Engine. In that sense, the "lack of context" shared by Gruber and the author is even more amazing than that of LLMs.