Pretty much every tech company has been investing in generative AI for the past two years, and Apple is rumored to finally bring new AI-based tools to iPhone users soon with iOS 18. However, the costs of operating such tools are quite high, and Meta has just confirmed that it will take years for the company to make money from AI.
Investments in Meta AI will take years to pay off
As Meta shared the results of its first fiscal quarter of 2024 on Wednesday, the company’s CEO Mark Zuckerberg spoke to investors about Meta AI, Meta’s new assistant that has now been integrated into apps such as WhatsApp and Instagram, and has also been added to Ray-Ban Meta smart glasses.
Zuckerberg seems quite excited about the future of Meta AI. However, the executive pointed out that it will take years for the company’s investments in generative AI to pay off. He noted that such experiments are a “very good long-term investment” and that the “initial signs are very positive.” Still, profiting from generative AI is “going to take several years.” (via The Verge)
Meta’s shares dropped more than 10% on Thursday after Zuckerberg confirmed that the company will spend even more money on AI. Meta recently announced that it intends to buy around 350,000 Nvidia H100 AI chips by the end of the year. Each of these chips costs between $20,000 and $40,000, which means that Meta will spend billions of dollars.
The costs behind generative AI are indeed quite high, as it requires a lot of processing power to train and operate such tools, not to mention the server infrastructure. Platforms like OpenAI’s ChatGPT charge users $20 to let them access more advanced tools like GPT-4 and DALL-E. Meta, on the other hand, doesn’t rule out monetizing its AI through advertising in the future.
What about Apple’s generative AI?
With all this in mind, where does Apple stand with its generative AI? We won’t have definitive answers until June at WWDC 2024, but we can speculate based on current rumors.
According to multiple sources, Apple has been exploring both on-device and server-based language models. While Bloomberg’s Mark Gurman recently reported that Apple’s new AI features will be powered entirely by an offline LLM, there are also reports about Apple negotiating with Google to adopt Gemini on its platforms.
Top comment by Graham Jones
I think the public still feels int he dark about AI. It’s a latest fashion to most people, and it’ll only become compelling when they see the evidence of AI helping them each day and feel its effects in a positive way. A lot of so-called AI features that have rolled out over the past few years have felt hidden to most people, and it’s most people you have to convince if you’re going to monetise it in any true sense. Until then Apple will probably offer it as a user facing feature and work out where it serves their purpose longer term. Apple can do online stuff well, but these systems have shown to be hungry on memory, so have Apple figured out how to get round that? Apple have always routinely updated several years of device models with similar features, and it would be a gamble to believe that people will simply buy this years iPhones based on the words AI, which are all the rage in the tech world and on Wall Street, but is simply the name of a sci-fi movie to most of the world.
If I had to guess, I’d say that the company will indeed use both offline and online models. Offline models have the disadvantage of not being as powerful and requiring more advanced chips, but they can also reduce server-related costs for Apple.
One possible scenario is for Apple to make offline AI available for free to run some basic tasks, while more advanced tools will run online and be paywalled – perhaps combined with an iCloud+ or Apple One subscription. This could help the company cover the costs of servers or even eventual deals with third parties such as Google.
The company already puts features like iCloud Private Relay under a paywall, so charging for a “Siri+” wouldn’t be a surprise.
What about you? Would you pay to use Apple’s generative AI? Let us know in the comments section below.
FTC: We use income earning auto affiliate links. More.
Comments