This ChatGPT Voice Update Previews What We Can Expect From Siri

A new ChatGPT voice update has made interacting with the chatbot far more flexible than it used to be, and I think is an example of one of the key things we can eventually expect from the new Siri.

While OpenAI’s change might seem like a relatively small one on the surface, I’m already finding that it completely transforms the experience of using ChatGPT …

Text to voice

When ChatGPT first launched, it was text only. Later, expanding the functionality to voice interactions had two impacts.

First, it made the chatbot seem even more like it was imitating a human being. That’s a whole topic of its own, but I’m with Dave Winer and John Gruber in thinking that this should not be an objective of a chatbot. (OpenAI did at least back down from a version which prioritized friendliness over accuracy.)

Second, it improved usability. It’s obviously faster and more convenient to simply speak to an AI than it is to type and read written responses, at least for certain types of interaction.

However, there are times when text and visual interaction is more appropriate, and while it was great to have the choice, it could be frustrating when you could get 95% of the job done with voice but then had to switch to text for things like asking ChatGPT to create an image.

This was especially true because the two modes were distinct, and it could feel like you were interacting with two entirely separate ChatGPT models – one voice, the other text.

Integrating the two

The latest update to ChatGPT integrates text and voice interactions.

You can now use ChatGPT Voice right inside chat—no separate mode needed. You can talk, watch answers appear, review earlier messages, and see visuals like images or maps in real time.

I suspect one of the things that is taking Apple longer than expected is providing exactly this type of integration. Effectively, the company is doing it the other way around: Siri launched as a voice-only service, and Apple Intelligence has now added text-based capabilities, such as summarizing emails.

OpenAI appears to be experiencing some of the challenges of pulling off this kind of integration. In my initial tests, it sometimes works extremely well and at other times not. For example, I’ve experienced it claiming to have shown me a map while it hasn’t in fact done so.

This is the future of the new Siri

If Siri is to deliver on all the promises Apple has made for it, then it will need to do exactly this and more.

We will need to be able to give it a voice instruction and have it not only give spoken responses but also manipulate text and images.

Additionally, it will need to act as an intelligent agent when interacting with our installed apps. We’ll give it an instruction like “Book the first flight tomorrow morning from Heathrow to Las Vegas,” and it will use whatever travel apps we have installed to carry out this task.

But Apple may end up subcontracting

Providing these capabilities is no small task. My jury is still very much out on how likely it is that Apple will catch up with not only where AI chatbots are today, but where they will be by the time the new Siri launches. I do think it’s very likely that Apple will have to subcontract AI capabilities to other companies.

Personally, I don’t care whether the technology used is homegrown by Apple or delegated to AI tech provided by OpenAI, Google, or anyone else. What I do care about is two things.

First, that Apple protects our privacy by customized interfaces with third-party chatbots in exactly the way it does with fallback to ChatGPT today. When we ask a question Siri can’t answer and it falls back to ChatGPT, our interactions are never used by OpenAI for training because that privacy protection is part of what Apple has agreed with the company.

Second, how quickly Apple is able to provide this intelligence to its devices. Next year is absolutely the far limit of how long we can possibly wait. If hitting that deadline means that what’s under the hood of the new Siri is a customized version of ChatGPT, Gemini, or something else entirely, that’s absolutely fine with me.

What are your views on this? Please share your thoughts in the comments.