Sarah Perez TechCrunch EXACTLY! Apple is taking baby steps because right now that is all we can count on to deliver successfully. This is a solid first effort by Apple focusing on real benefits to users (ex. Siri). My two cents… Today's AI LLMs are mostly glorified demos for the really interesting applications. They are turning into a commodity because they're overfunded by FOMO-driven VCs who don't truly understand the technology limitations that drive real application requirements. Hallucinations are a real problem and there is no fundamental way to get rid of them. The expectations of customers are so much higher than what can be delivered today. Remember how everyone ran to fund self-driving cars in 2012-2014?!? It was the NEXT big thing. What do we have 10-12 years later? Think about how many years have past for those efforts to be perfected (and they still aren't yet). We haven't even seen a self-driving car business work yet either. How many billions were spent on that next big thing before reality set in?! AI is VERY similar. AND worth noting. There are many other approaches to AI that actually work reliably today. While they may not have magical LLM-like capabilities, they are solving very valuable problems and have businesses that work well. It will take many years to get to the AI capabilities we expect, unlike self driving cars, we can start using the right AI today!
Alon Saban
1mo
Apple's reputation for innovation precedes itself. Time and again, they've set the bar high, inspiring waves of imitation and establishing new industry benchmarks. What sets Apple apart, in my opinion, is their commitment to letting you experience their products firsthand. When you interact with an Apple device, it's like touching the future. I felt this connection with my first MacBook, my iPod, my iPod Touch, my iPhone, and the list goes on. You don't truly grasp the essence until you've tried an Apple product—the seamless convergence of technology and user experience. it's a mindset. Admittedly, Apple's approach to artificial intelligence may seem somewhat outdated. Their presentations lack the glitz and glamour typically associated with AI advancements. However, as history has shown with every other Apple technology, you inevitably find yourself enchanted and comprehending their magic once more. In this narrative, OpenAI takes a backseat. What truly matters are privacy, on-device intelligence, and personalized assistance tailored specifically for you. While Apple does utilize OpenAI technology, they do so in a manner that aligns with their values and philosophy, ensuring they stay true to their vision.
Also because everything is not a generative problem; it's a multi disciplinary AI problem. If AI has to mimic behavior then we need to classify human behavior into x buckets - our memories (that video of swimming when we were 7yrs old) is a knowledge graph - our ability to infer - what do you think I'm going to..... (You're thinking say or do) thats more like a neural network / graph netural network - let's make a picture about a meeting with everyone in the room - that is a RAG / graph based RAG with generative AI to produce it. The key is that the problem sets above - less then 10-15% is a generative behavior.
Emmanuel Oluga
1mo
It's an interesting take, but it conflates two orientations. OpenAI IS already FOMO-over-indexed. Apple's choice to go with them is most definitely FOMO as well. LLMs WILL become a commodity, and not even the best take on AI anyway, but that's a discussion for another time. Their presence here will barely make any dent that should have been made many years ago. To your point about the Baby Steps that lead to real benefits to users (ex. Siri), this is not it either. If Apple truly wanted baby steps, Apps would have been able to communicate with each other long ago, and still be subject to the edifice of the Apple Privacy stance without any shortcomings. We've had smart Apps in smartphones for over a decade because Apple and Android have refused to allow innovation and "feign defense of privacy" instead Capitalism is my lifeblood, but the type practiced by Apple is straight-up predatory. I am forever grateful to you for your radiant transformation of tech, but this is not the take, especially knowing the egregious shortcomings of openai thus far, which are antithetical to protecting user privacy. First, I'd love to see Apple enable entry points for rejuvenating solutions like Ledger Stax to thrive alongside their ecosystem.
🤔 Steve Jennis
1mo
I’m disappointed that Apple lost its nerve to FOMO rather than continuing with its heritage of getting it right, rather than first. After Vision Pro rebranding VR as spatial computing, and now AI being rebranded as Apple Intelligence this looks like a worrying trend of MVP launches to catch a hype wave, rather than solid products that provide privacy and security as well as gimmicks and glitz. Maybe now the 🍏 isn’t yet ripe?
Jiri Kram
1mo
Exactly Tony Fadell. For me it started in a gym. When few months ago Siri started to talk to me over iPods. So I tried “play some music”. She started playing music. Then I said “play Metallica”. It played Metallica. Then I said “play Metallica Master of Puppets” again it played this song. Then I decided how much Siri understand so I said “Hey Siri play Metallica black album”. When I heard first accords of Entersandman, I realised the potential. Same as they trained Siri to recognise unofficial name of Metallica 1991 album it can connect to any other LLM and Siri will become my most used AI. And they didn’t disappoint.
Puneet Jindal
1mo
Tony Fadell Agree with you that it made sense for Apple to go with integration of AI with user first approach and applied similar to Rule 1 of machine learning (https://developers.google.com/machine-learning/guides/rules-of-ml#rule_1_don%E2%80%99t_be_afraid_to_launch_a_product_without_machine_learning) and later improve and build it with feedback, however its sad that you literally reduced down the efforts of all those FOMO driven VCs because to achieve Apple intelligence has been made possible by all those efforts put in the past including those of post and pre transformers era whether it is GPTs or self driving tech which laid the foundation. Definitely we can say that we under estimate the self driving definition of done but thats how technology evolves over time, isn't it?
Ying Liu
1mo
"There are many other approaches to AI that actually work reliably today. While they may not have magical LLM-like capabilities, they are solving very valuable problems and have businesses that work well" I cannot agree more. Everyone is excited about AGI, but do we really need AGI for average users in most scenarios? We need more low power and device-based solutions, vs. cloud and LLM.
MSc, DIC, MBCS | Xbox Accessories Firmware | Microsoft | ex-Apple | ex-Palm
1moExactomondo. I am surprised with all the articles in places that should maybe know better (like the FT) with a "how is Apple going to catch up" vibe. A far better approach is "how will BigFruit leverage models to deliver better products". I was looking at getting an ONNX model implemented into an embedded system. Since it's ARM based, I looked at ARM NN which wants "4GB RAM and 1 GB Flash" as it's minimum operating environment. Crazy. Pushing everything to the cloud is not a solution, sometimes I think we are dragging computing back to the 1960s with a reliance on Big Iron, this time in the cloud.