Apple has created its own generative AI tools to compete with large language model (LLM) services such as ChatGPT or Google Bard, Bloomberg reported on Wednesday.
Bloomberg (which also today reported plans for Apple to buy Disney), says Apple has crafted its own internal framework to build LLM models.
- That model, “Ajax,” has been used to build a chatbot service apparently internally referred to as “Apple GPT.”
- Run on Google Cloud, the latter tool reportedly has features similar to popular LLM services.
- That means it can answer questions based on data it has been trained on, summarize text, and so forth.
- The current version runs as a web application and is being used internally for product prototyping.
“The chatbot app was created as an experiment at the end of last year by a tiny engineering team,” Bloomberg says, adding, “Any output from it can’t be used to develop features bound for customers.”
No go-to-market plan…yet
Apple executives don’t yet have a clear strategy to bring this to market, but they will be feeling some pressure. One of the most successful product launches in history, generative AI has become the big story of 2023 on the back of ChatGPT. While this powerful technology has its flaws and has seen governments worldwide launch investigations exploring ethical use of the tech, most software developers are building tools that use it. Microsoft with Bing/OpenAI, Google with Bard, and Meta and Qualcomm with Llama 2.
Adobe has introduced its own highly focused implementations of generative AI within its creative products, for example.
While Apple does have AI baked inside all of its products, it didn’t really have a stake in this game. Apple CEO Tim Cook has instead warned that there are “issues” that need to be resolved around useing the tech. He also said Apple will add AI to more of its products “on a very thoughtful basis.”
Bloomberg now claims Apple aims to make a significant AI-related announcement in 2024.
The Trojan wars
It is interesting to reflect on the name. As Siri may already be able to tell you, Ajax is a classical Greek hero of the Trojan Wars featured in the Illiad. A great warrior of immense courage, Ajax is described as strong, powerful, and very smart.
The choice of name is at least a departure from California holiday destinations and may reflect the scale of what Apple wants to achieve. Because despite seeming late to the party, Apple has major resources to bring to bear to make this technology a success.
One of the biggest advantages it has will be the search data the company has gathered in support of Spotlight. In addition, the company will possess high quality, yet anonymized, information pertaining to search intent.
But, perhaps even more importantly, any LLM tech Apple does eventually choose to deploy will have access to all of a person’s on-device information and access to the machine intelligence-supporting Neural Engine that sits inside every device.
In other words, the LLM will be able to access a person’s personal information to generate accurate and relevant responses and will be able to do so without any personal information whatsoever leaving the device or being shared with Apple.
All it needs is to figure out what tasks people need the tech to achieve, and then create relevant frameworks. (Generative AI-based image editing in Photos comes to mind, for example.)
A boost to enterprise?
Privacy isn’t just an advantage in consumer makets; it also matters within the enterprise. Anxious to protect company data, major enterprises including Apple, Samsung, and others have banned employees from using ChatGPT or GitHub Copilot internally.
The desire to use these tools exists, but not at the cost of enterprise privacy. Within the context of Apple’s growing status in enterprise IT, the eventual introduction of LLM services that can deliver powerful results while also having privacy protection built in means the company will be able to provide tools enterprise employees might be permitted to use.
Not only this, but those tools could end up displaying a degree of personal contextual relevance that isn’t available elsewhere — without sharing key personal data with others.
So, there’s a lot of optimism; it is, after all, not the first time Apple has appeared to be late to a party and then delivered a better experience than available elsewhere.
This optimism was reflected swiftly by investors. While warning that the next iPhone may not ship until October, Bank of America raised its Apple target to $210 per share from $190. Credit Suisse raised its Apple price target $20 to $220 per share.
The news drove Apple shares to reach $197.32 before settling to around $193 per share.
Please follow me on Mastodon, or join me in the AppleHolic’s bar & grill and Apple Discussions groups on MeWe.
Copyright © 2023 IDG Communications, Inc.
This story originally appeared on Computerworld
