Faced with the rapid development of ChatGPT, how does Apple respond?

Before the ChatGPT came out, it was still a long-standing practice for global technology companies to wait for Apple's answer.

Whether it was the crossover self-driving cars or the VR/AR devices that were being called the next generation of mobile hardware, practitioners and consumers around the world were often somewhat "superstitious" about the final answer that Apple would give down the road. But this absolute dominant innovation authority is being pried by ChatGPT.

As the hottest direction of artificial intelligence, ChatGPT was launched by a startup company, OpenAI, and has taken the world by storm. Even with the backing of Microsoft, the old gold mine, this non-major tech innovation path is still causing people to start re-examining the innovation capabilities of traditional tech giants.

Google became the first major Internet company to be hit. By investing in OpenAI, Microsoft integrated ChatGPT into its Bing business, launching an attack on Google's dominant position in search. Subsequently, Google announced several AI-centric business redo plans to play defense.

Alarm bells are ringing and an arms race around AI has become almost a consensus among global majors. After Google defensively launched Bard, Meta launched a chatbot model called LLaMA in February, and Amazon launched a service called Bedrock in mid-April.

As the creator of the last iPhone moment, Apple is almost untraceable in this GPT wave. Faced with the surging wave of artificial intelligence, the former leader has surprisingly become a fringe player.

Although Apple has always had the success story of post-emergence, this time, facing the new era, Apple is also questioned by the elephant dancing. Whether it can find its own AI path determines whether Apple can continue to sit on the iron throne of the global technology powerhouse.

I. Dilemma of Siri under Apple's rule

Fixing and improving on Siri is by far the biggest signature move Apple has made in response to ChatGPT.

The latest public news shows that Apple's software engineers are proposing to integrate the machine learning technology behind chatbots like ChatGPT into the company's Siri voice assistant. Considering that Apple has been very careful with Siri's development over the past 12 years, there is no guarantee that executives will approve these proposed changes.

Back in 2011, Apple purchased the technology from SRI International and introduced Siri, a virtual assistant, about a year later. When it debuted, Siri, which is based on voice recognition and NLP technology, was first featured on the iPhone 4S.

In a test conducted in 2019, Siri answered 800 questions correctly 83 percent of the time, a 4 percentage point improvement year-over-year. In this virtual assistant war, Siri outperformed Alexa but not as well as Google Assistant, but now all three have undoubtedly squandered their previous leadership in the face of a new generation of AI technology, compared to ChatGPT.

Compared with voice assistants of the same era, Siri is lucky that Apple has not yet given up on it, but is ready to make it a core force to catch up with ChatGPT through transformation.

But this transformation will not make Apple stand on the "shoulders of giants", but rather because of the legacy of the previous generation, even more difficult than re-engineering.

A former Apple engineer who worked on Siri said in an interview that Siri was built on "clunky code that took weeks to update basic functionality" and that its "clunky design" made it difficult for engineers to add new features.

This means that to catch up with ChatGPT through Siri, Apple may need to rebuild the entire back-end technology in terms of technical implementation. As a result, some Apple engineers are proposing to connect it directly to ChatGPT technology, but the chances of that being adopted are slim at this point.

For Apple, as a long-time leader of the old era of the giant, Siri's technical path problem is just the tip of the iceberg of the historical legacy. In the face of the new impact, Apple's difficulties come more from the corporate culture genes, organizational structure and other more from the ills of large enterprises.

On the one hand, in the corporate genes, Apple's system for data usage makes it challenging to build large language models. Apple executives, led by Cook, have been sounding the alarm on data privacy and security, while constantly tightening their ecosystem to restrict the flow of personal data.

In contrast, the big-model direction represented by ChatGPT requires exactly the kind of massive intake of information needed to mimic what humans might create, often from opaque sources. The difference in stance on data use makes it necessary for Apple to first compromise on its corporate culture if it wants to catch up with ChatGPT.

But that won't be easy. At previous developer conferences, Apple CEO Tim Cook has stated that the company's stance on data privacy is the opposite of Google's. Apple analysts believe that the focus on privacy may slow down Apple's progress in building some of its products, but it can be exchanged for consumer trust.

On the other hand in terms of talent organization, Apple has not demonstrated a stronger appeal. Over the past eight years, Apple has lagged behind other large tech companies in hiring AI researchers, and a series of recent senior-level departures could also hinder its AI business.

Last year, Apple lost Goodfellow, its director of machine learning and a prominent AI researcher, because he was unhappy with the company's policy of returning to work in its offices. Apple's seeming reluctance to be more flexible in retaining Goodfellow meant that the business of artificial intelligence, at the time, was not a priority within Apple.

Apple also recently lost Michael Abbott, head of cloud computing, a key division for building and deploying AI services. In terms of talent acquisition, Apple still has 342 job openings in AI-related areas such as deep learning and natural language processing.

To add insult to injury, at the same time Apple was undergoing its Siri makeover, OpenAI introduced its next-generation AI engine, GPT-4, which enables ChatGPT to make more advanced responses. This further prompted the market to start re-evaluating the way big companies like Apple, Meta and Amazon develop their technologies.

II. The new AI bet beyond the big models

To challenge Microsoft's leadership in AI, Apple clearly needs to reconsider the new way it implements AI.

By scanning recent trends in AI talent turnover at large tech companies, analytics firm Glass.ai estimates that Apple appears to have avoided entering the ChatGPT race in the past six months by continuing to hire visual recognition experts rather than big language modeling technologies.

As another area of artificial intelligence, visual recognition is more aligned with Apple's work on upcoming AR devices in the future. This means that Apple will combine its historical strengths in building technologies used by creative professionals and consuming content such as music and movies, with a potentially greater focus on visual and audio content in the area of generative AI.

This strategy, while not as compelling as ChatGPT's automated generation of large amounts of text, is more relevant to Apple's strategy of devices as infotainment portals.

While Apple is a late starter in new areas, there are precedents for successful improvements in the past, such as the iPhone and AirPods, but these are the areas of hardware in which Apple excels. Online services are relatively more challenging, and Apple's performance in this area has been controversial, with Apple Maps, iTunes Ping and early cloud woes.

One reason for this is Apple's centralized organizational structure and designer-led culture, which has built its work model primarily around product development. Creating large-scale online services that can get to market quickly requires more of a decentralized organizational structure and engineering-driven culture like Google's.

If visual recognition is doing AI exploration for unknown hardware products, then putting AI on the ground quickly in health may be another important direction Apple is currently taking for AI.

On the Q1 2023 earnings call, Cook said, "AI is a horizontal technology, not a vertical technology. As such, it will impact every product and every service we have." This means that Apple's efforts in AI will likely focus on improving products and services, rather than creating new standalone AI products.

And of all the products and services that could be improved and optimized, health could be the center of Apple's storm for AI applications. Previously, in February, Cook told investors that AI would be Apple's "primary focus" and said the company had applied it to features like collision detection on the Apple Watch and iPhone.

Two months later, an April 26 report from Bloomberg revealed that Apple is developing an artificial intelligence-powered health coaching service and new technology to track mood, its latest attempt to target users with health and wellness features.

The new coaching service, codenamed Quartz, aims to keep users motivated to exercise, improve their eating habits and help them sleep better, according to people familiar with the matter. Their idea is to use artificial intelligence and data from the Apple Watch to make recommendations and create tailored coaching programs for specific users.

The project is being driven by several Apple teams, including its Health, Siri and Ai teams, as well as its Services division. Also included in this program is the expansion of the Health app to the iPad, as well as features that can help users with poor vision.

In the short term, Apple plans to debut an iPad version of its iPhone health app. The change will allow users to view ECG results and other health data in a larger format, and is also planned for later in the year as part of the next-generation version of iPadOS.

Tools for tracking mood and managing vision conditions will also be added to the health app this year. The initial version will allow users to record their moods, answer questions about their day and compare results over time.

In the future, Apple hopes the iPhone will be able to use algorithms to determine a user's mood by their voice, what words they type, and other data on their device. The goal is to introduce new iPad apps, as well as tools to manage emotions and visuals, at its annual conference this June.

The company's upcoming mixed reality headset display device, also set to debut in June, will also play a role in the health and wellness strategy. A feature will allow users to meditate while wearing the device, and an optimized version of Apple's Fitness+ workout service is in development.

If its AR devices are a big success, Apple's bet on visual artificial intelligence will clearly be a successful gamble. But the absence of standalone products on big models, the shakeup of AI talent executives and the corporate genetic issues under Siri all make Apple today more risky than ever.