If not for the Silicon Valley Bank (SVB) collapse last week, almost every conversation in the tech industry seems to revolve around AI and chatbots. In the last few days, Microsoft-backed OpenAI has released a new language model called GPT-4. Competitor Anthropic has released a chatbot, Claude. Google said it is integrating AI into workspace tools like Gmail and Docs. Microsoft Bing is gaining traction with its chatbot-enabled search.
Last month, the Cupertino-based company hosted an internal event focused on AI and large language models. According to a New York Times report, many teams, including those working on Siri, regularly test “speech generation concepts.”
People have long complained that Siri doesn’t understand queries (including mine). Siri (and other assistants like Alexa and the Google Assistant) can’t understand the different accents and phonetics of people in different parts of the world, even if they speak the same language.
ChatGPT and the newfound fame of text-based search have made it easier for people to interact with various AI models. But for now, the only way to chat with Apple’s AI assistant, Siri, is to enable the feature in Accessibility Settings.
In an interview with the NYT, John Burke, a former Apple engineer who worked on Siri, said Apple’s Assistant has been slow to develop because of “jerky code” that makes even basic feature updates difficult to push. He also mentioned that Siri has a large database stack with many words. So when engineers needed to add features or phrases, the database had to be rebuilt. The process reportedly took up to six weeks.
If Apple wants to create its own language model or adopt an existing model. But like Google and Microsoft, the Tim Cook-led company doesn’t want to stop at offering Siri-powered chatbots. Apple has long prided itself on being an ally of artists and creators, and hopes to apply language model advances to these areas.
The company has been using AI-powered capabilities for some time, but it may not be obvious at first. These include Better Suggestions on the Keyboard, Processing on Photos, Face ID to Unlock Masks, Separating Objects from the System-Wide Background, Hand Washing and Crash Detection on the Apple Watch, and most recently, Apple Music. Includes karaoke function. However, none of them managed to attract attention as a chatbot.