Just 19 months ago, global markets bottomed out, with the US benchmark S&P 500 index grazing the 3,577 mark on Oct 12, 2022. The market is up over 50% since. The main driver of the bull run has been the advent of generative artificial intelligence (Gen AI).
Six weeks after the markets bottomed out, start-up OpenAI, in which Microsoft has a 49% economic interest, unveiled ChatGPT, a chatbot built on foundational large language models (LLMs) that understands not only the words and syntax that we use in our everyday lives but also their nuance and context. ChatGPT can craft email, generate computer code and even write poems. The world has not been the same since.
Think of Gen AI the way you think of calculators. You punch in a bunch of numbers and it can churn out multiplications, divisions, percentages and even complex calculus which even math gurus cannot do quickly. When you ask a question, the LLM calculates how to respond to you based on the relationship of various words to each other that it was trained on using hundreds of terabytes of data from books, newspapers, magazines, blog posts and websites. A terabyte is 1,000 gigabyte — equivalent to the words in thousands of Encyclopaedia Britannicas.
Chatbots learn just like how kids learn to communicate: They hear words from their parents, siblings, friends or on TV and they put those words together and spit them out. Early words and sentences are jumbled, but as children grow, they become more eloquent.
That is exactly what is happening with ChatGPT and other AI chatbots. There is a race to train the LLMs on as much high-quality data as possible. The next stage is getting the algorithms to use all their training to make basic inferences. That is what is unfolding now. Chatbots can do tasks like getting the cheapest airline ticket for a trip, book a hotel room or pay bills. They can also execute voice commands on your smartphone by saying: “Hey Siri, pay $12 to the dry cleaning store. Use my Visa card” or “Hey Alexa, turn off the lights in the living room, thank you.” Last year, I bought a coffee machine that works with a smartphone app. My machine will soon work with a voice-activated chatbot. Instead of struggling with the app, I will be able to tell the machine to make me a macchiato just as I am finishing up on the treadmill at the gym.
Since ChatGPT was first unveiled in November 2022, a handful of tech companies have seen the race to embrace AI dramatically transform them. The biggest winner has been the chip powerhouse Nvidia, whose stock is up a whopping 960% since then. Another beneficiary is Microsoft, which has a 49% economic interest in ChatGPT maker OpenAI and a stake in French AI start-up Mistral. Microsoft also recently gutted another American AI start-up, Inflection AI, by hiring away almost all of its senior staff.
Others like search engine giant Google’s parent Alphabet, dominant e-commerce and cloud computing firm Amazon.com and social media supremo Meta Platforms, which are spending tens of billions of dollars buying AI chips from Nvidia every year, have enjoyed the pixie dust of AI helping them sparkle over the past 18 months.
See also: UBS-Credit Suisse integration opens up new tech for bigger plans
Apple Intelligence
Apple, the world’s largest company until five months ago, saw its own fortunes plummet because it seemingly had no AI strategy of its own to boast about. It was overtaken by Microsoft in mid-January as the world’s most valuable company and two weeks ago, even Nvidia briefly surpassed Apple before retreating back to third place. The narrative was that while hyperscalers like Microsoft, Amazon, Google, Meta and others spent more than US$90 billion ($121.5 billion) on AI-related chips last year, Apple spent almost nothing and as such it should be treated as a laggard in the AI era.
On June 10, at its annual Worldwide Developers Conference, Apple finally unveiled its AI strategy.
It branded its offering as Apple Intelligence, a suite of AI-based capabilities that promises to deliver personalised AI services while keeping your sensitive data secure. Apple wants to use our private data to help AI do tasks for us. But unlike Google, Meta, Microsoft and others who suck your data, sell advertising against it or sell it to data brokers willing to pay the highest price, Apple is promising to do this without sacrificing privacy. Because Apple makes far more money from hardware and services than from ads, it is less incentivised to collect personal data. That has helped it position the iPhone as a private device.
See also: Google arguments draw scepticism from judge in ad tech case
The iPhone maker has built a new way to handle sensitive data in the cloud. Apple’s new “Private Cloud Compute” eliminates persistent data storage and removes a remote shell that could allow access to user data. Data that is exchanged with other cloud services will be encrypted and then deleted afterward.
So, what’s new? Apple presented its AI offerings that powered apps with new capabilities. Siri, its much-derided 13-year-old voice assistant, is now more capable thanks to help from AI. Among other things rolled out was a new version of Mail that can generate complex email responses. The new AI-powered Safari can now summarise web information for you. Want a quick summary of what’s happening in your favourite sports or in your hometown? Safari will give it to you to browse with your morning coffee.
Unlike other companies developing Gen AI, Apple Intelligence is not based on LLMs but learning “personal relevance” by pulling data from iMessage, email, Siri and photos or data it has accumulated from its huge iPhone installed base, notes Edison Lee, analyst at Jefferies & Co in Hong Kong. Apple’s AI solution is cheaper and more useful. Unlike OpenAI, Meta, Microsoft and Musk’s xAI which use expensive Nvidia GPUs, Apple uses its own graphics chips for training and inferencing of data. “Apple can now use AI services to differentiate between high- and low-end models, motivating iPhone users to upgrade to the latest Pro or Pro Max models, boosting volumes and average selling prices,” Lee notes. “Android may find it hard to match.”
Apple’s strength is its integrated approach with its own silicon chip, hardware, software and services. Android’s problem, says Lee, is it does not design its own silicon. Moreover, because no single Android phone maker has a large enough user base, it is hard for them to develop competitive personalised AI services the way Apple can. “Apple Intelligence will further increase the stickiness of the iOS ecosystem,” Lee says.
Edge AI is the key
The key to Apple Intelligence is Edge AI, a new computing segment that addresses the need to process data closer to where it is generated or where things and people connect with the digital world without having to constantly rely on cloud infrastructure. Edge AI deploys algorithms and AI models directly on sensors on smartphones or tablets, which enable real-time data processing and analysis. Because you and I will need to access AI from our wearable devices, like smartphones, smart watches or ear pods, which account for most of the data we generate or consume every day, it is important to process that data closest to where it is being used or generated.
Think of the way most restaurants work. They could source great dishes from a cloud kitchen miles away, but any savings would be useless because it would take too much time before the food arrives on your table. So, restaurants cook food on-site. Just as people want their food now and hot off-the-oven in a restaurant, they would rather have data processed on their devices than have it sent to some data centre cloud server elsewhere for processing.
Doing most of AI computing on the devices also helps Apple’s focus on privacy. Instead of sending every query to third-party servers, Apple will ask permission from their iPhone or iPad users if they want to send the query to ChatGPT. Apple will divert the query only if its users want it to.
Sink your teeth into in-depth insights from our contributors, and dive into financial and economic trends
Apple is using a foundation model trained for each specific task like summarising, proofreading, or generating email replies. Apple is further fine-tuning its small language model (SLM) using adapters (a small collection of model weights) overlayed on the SLM to further train for specific tasks. The end goal of Apple’s on-device solution is to shorten the time from a prompt to a response on the device as much as possible.
Apple knows that while most people around the world have read or heard about AI, they have not yet engaged with Gen AI. “To most people, AI doesn’t mean anything,” says Neil Cybart, who writes the Above Avalon blog about all things Apple. As the world’s biggest consumer electronics firm, Apple is focused on building AI functionalities into its products. For most people, their smartphone, laptop, tablet or smartwatch will be the first AI-powered thing they will interact with. As such, Apple’s focus has been personalising these and making them user-friendly.
To be sure, Apple’s ecosystem is huge. There are about 2.4 billion Apple devices — iPhones, Macs, iPads and Apple Watches — currently using iOS, Apple’s operating system. More than 1.7 billion people (out of 8.1 billion worldwide) use at least one Apple product. Of those, 1.5 billion use an iPhone. If you want to quickly reach the top 20% of the people on earth, you need to go through Apple. That makes Apple the world’s biggest and most prosperous gatekeeper. Everyone wants access to Apple’s affluent customers. Google pays US$20 billion every year to Apple so that iPhone users can use its search engine. Google generates about US$100 billion in additional advertising revenues from Apple customers. In 2023, Google had US$239 billion in total ads revenues. So, about 40% of Google ads are coming from Apple users. Google also pays Samsung and other smartphone makers, but those payments are a small portion of what Apple receives.
It is unclear how the OpenAI-Apple partnership will work. Will Apple pay OpenAI tens of billions of dollars a year for accessing its state-of-art chatbot GPT4, or will OpenAI eventually end up paying Apple billions to get access to the world’s most prosperous customers — or 20% of population that use high-end Apple devices — to compete head-to-head with Google, whose lower-end Android devices cater to the bottom 80% of world’s population? While Google has Apple as its main search partner, OpenAI — which is also disrupting Google’s search business with its chatbot to answer queries — has Apple as its anchor AI partner.
On June 12, two days after it unveiled its AI strategy, Apple briefly pipped Microsoft to retake the position of the world’s most valuable company, only to see Microsoft narrowly regain it in the closing minutes of the trading day.
The emergence of Apple as the tech giant best positioned to take AI to the masses is unsettling Big Tech peers. EV pioneer Tesla’s CEO Elon Musk, whose xAI is also hoping to target AI for consumers, has fired the first shot. Employees of Tesla, SpaceX and his other companies will be banned from bringing iPhones to work. OpenAI’s new partnership with Apple could be problematic for Microsoft, which has invested US$13 billion and has a 49% economic interest in OpenAI but little control over the firm or its charismatic CEO Sam Altman. Microsoft is also eyeing consumer AI. Two months ago, it hired Mustafa Suleyman, a former top executive of Google’s DeepMind as its consumer AI chief.
Apple also has to be mindful of its relationship with Google, which pays it US$20 billion a year. By directing search queries to ChatGPT, Apple is taking away traffic from Google which needs it to generate more ads. Google could stop paying Apple any money, but that would be at the cost of fewer ads.
It is still early days in the battle for AI supremacy, but Big Tech alliances could see fundamental changes over the next few years.
Assif Shameen is a technology and business writer based in North America
