Is Apple developing its own large-scale language model and chatbot AI 'Apple GPT'?
Bloomberg reporter Mark Garman, who is known for leaking a lot of information related to Apple, said that while Apple is building its own large-scale language model, it is called ' Apple GPT ' internally. It reports that it is developing a chatbot AI.
Apple Preps Ajax Generative AI, 'Apple GPT' to Rival OpenAI and Google - Bloomberg
Apple is testing an AI chatbot but has no idea what to do with it - The Verge
https://www.theverge.com/2023/7/19/23800430/apple-gpt-ai-chatbot-generative-ai
Since Google announced a deep learning model for natural language processing called ' Transformer ' in 2017, AI companies such as OpenAI, which develops ChatGPT, and major tech companies such as Google and Meta have all been promoting the development of AI. increase. Google announced large-scale language models `` LaMDA '' and `` PaLM 2 '' and chatbot AI `` Bard '' based on them, and Meta announced large-scale language model `` Llama 2 '' that can be used for free for commercial use. . However, although Apple is developing Siri, an assistant AI to be installed on the iPhone and Mac, it is pointed out that it may have missed the wave of AI development since the appearance of Transformer.
Pointed out that Apple's cautiousness and malfunction of the development team are hindering the efforts of the voice assistant 'Siri' and AI, and there are concerns about missing the new era of AI-GIGAZINE
Apple was reluctant to develop AI because CEO Tim Cook was cautious about AI. Cook CEO said, ``Although AI technology has potential, there are still many problems to be solved,'' and although he acknowledged that Apple will incorporate AI into its products in the future, ``It is It will be done very carefully,' he said.
However, Mr. Garman said that he heard from an anonymous official that AI development was one of the important efforts for Apple at the time of writing, and that multiple teams are collaborating on confidential projects. . The AI development effort is led by John Janandrea, head of Apple's machine learning and AI team, and Craig Federighi, senior vice president of software engineering.
Garman reports that Apple is building a framework called 'Ajax' for developing large-scale language models. This 'Ajax' was first built in 2022 to integrate Apple's machine learning development, and was originally developed based on Google's machine learning framework ' Jax '. Ajax is already running on Google Cloud, and it seems that AI technology based on Ajax is introduced to Siri and Apple Maps to improve it.
In addition, based on this large-scale language model developed on Ajax, a chatbot AI has also been developed, and some engineers call it 'Apple GPT', Garman reports.
Apple GPT was developed experimentally by a small engineering team at the end of 2022. Apple initially stopped deploying Apple GPT internally due to security concerns about generated AI, but it seems that more employees have since been involved in development. Apple GPT is basically a reproduction of Bard and ChatGPT, does not include new functions or technologies, and can be accessed internally as a web application. However, as it is a confidential project, access to Apple GPT requires special approval.
In addition, Apple GPT is only used by some internal employees, it is not used for consumer product development, and Apple GPT is not planned to be released for consumers, Garman said. I'm here.
Related Posts:
in Software, Posted by log1i_yk