Apple might be the next company to add a ChatGPT-style chatbot to its Support app, likely as a way to automatically handle cases covered in its documentation and hand off only the trickier ones to a human. All things considered, this is a good move.
A ‘Support Assistant’ potentially powered by a third-party model
The Support app already has a live chat feature that connects users to actual humans, but wait times can get a bit unpredictable depending on the time of day, or after peak seasons like Christmas and product launches.
As reported by MacRumors, Apple is developing an AI-based “Support Assistant” feature for its Support app.
According to strings found in the Support app, the assistant “uses generative models,” which is broad enough that it could mean that it probably won’t be using Apple’s own models, at least not yet.
Enter RAG
Assuming Apple doesn’t use its own models, what it could do is leverage the same kind of B2B infrastructure already offered by companies like IBM, NVIDIA, and Salesforce, all of which have built solid enterprise tools using a technique called RAG (Retrieval-Augmented Generation).
The basic idea behind RAG is simple: instead of relying solely on the tokens that a language model was trained on, you also feed it with specific, real-time information, like the entirety of Apple’s support documentation in every language, so that it can formulate the answer to a prompt.
Even better, RAG is structured in a way that the system intelligently picks out just the relevant documents, or even segments of documents that it needs to answer each specific prompt, and builds its answer on top of that sliver of relevant information rather than having to read through everything with every request.
Interestingly, the report also mentions users being able to upload images or documents. That could allow the AI to analyze a photo of a cracked screen or parse an error message from a screenshot, capabilities that would pair naturally with a RAG-based backend.
... continue reading