Update: Our OpenAI event live blog is up and running. Follow along for all the latest news and announcements as they happen.
OpenAI is joining the likes of Google, Apple and Microsoft in hosting an event to promote a new product and it’s got a hype similar to anything held by the iPhone maker.
So what can we expect from OpenAI’s first proper public announcement event? Probably not what you expect to see as the company begins to focus more on product than model.
The rumour was that we are going to see some form of new search engine and possibly GPT-5, but none of that is going to happen according to CEO Sam Altman.
He said on X that the announcement is “not gpt-5, not a search engine, but we’ve been hard at work on some new stuff we think people will love! feels like magic to me.”
The rumour mill leans heavily on the fact that we’re going to get a voice assistant, and this isn’t a voice assistant like Siri or Alexa, this will be closer to Samantha from the movie "Her" — a true end-to-end conversational AI.
What can we expect from the OpenAI event?
I agree with the rumor mill that a voice assistant of some description is the most likely major announcement at the OpenAI event.
However, creating a true voice assistant will involve some significantly upgraded models including improved speech recognition and voice analysis. This will likely mean a new version of OpenAIs already powerful Whispering transcription model.
We may also possibly get agent like behaviour for the new assistant. This is where the AI can go off and perform actions on its own on your behalf around the wider open web.
These alternative models and potential agents could find their way into ChatGPT Plus, the premium plan for OpenAI’s flagship product.
If we get a big upgrade for ChatGPT Plus it will likely mean an upgrade for the free version, finally bringing in GPT-4 and DALL-E.
A deal with Apple?
One of the most noteworthy announcements from today's event would be a deal struck with Apple giving the iPhone maker the rights to use OpenAI's tech in its own products. Such an agreement has been hinted at by Bloomberg’s Mark Gurman and would presumably be the foundation for Siri 2.0. In fact, Apple's voice assistant could be a rebranded ChatGPT — much as Microsoft has done with Windows Copilot.
Given that Apple's own WWDC is only a month away, it could be the case any such agreement is held until then. But if such a deal is in the making, perhaps today's event will give us a very small teaser.
What form will this magical voice assistant take?
In the movie "Her", the AI character Samantha is designed to adapt and grow through her interactions with humans. Over time Samantha develops self-awareness, emotional depth, and the ability to form meaningful connections.
We have seen hints of OpenAI leaning in that direction. ChatGPT can now remember what you typed and use that in future conversations, and if you’ve ever interacted with the voice agent in the ChatGPT app it includes human like pauses and inflections making it sound more emotive.
I don’t think for a minute, we’re going to see anything on the scale of Samantha. But, if OpenAI has created an improved end-to-end voice AI, able to act on your behalf and be integrated into other device — it will be a “magic” moment.
The biggest change will be a move to speech-to-speech. Currently ChatGPT Voice converts your speech to text and then takes that text, sends it to the AI model, gets text back and converts that to speech. This creates a lag that doesn’t work well for conversation.
Unlike Siri or even Gemini where you interact by asking a question and wait, hoping it has been trained on or programmed to have the answer, with a new true voice assistant you’ll just have a natural, human-like conversation.
What about the agents?
Agents are the next big trend in artificial intelligence. They are mini AI models controlled by the main model like GPT-4, but capable of handling tasks on their own.
For example you could tell ChatGPT “it’s my wife’s birthday and I forgot” and it could go off, find a present from what you’ve said about her in the past, order the present and arrange delivery along with messaging your wife.
An example of these “swarms” of agents can be seen in the AI developer platform Devin, where you tell it what to make and it goes off and performs all the actions it needs to achieve the goal from web browsing to downloading images.
What else might we see?
We might get more Sora videos and an indication of when it will be available to the public. We might also get the first idea of how well the Voice Engine ElevenLabs alternative actually works.
The focus is going to be on products rather than underlying models. We are entering the commercial era of AI where those impressive models have to earn their keep.
That isn’t to say we won’t see new models. Altman has already said GPT-5 will be a significant improvement on GPT-4 and that he’ll break the bank to develop a superintelligent AI.
This event feels more like the AI Lab entering the commercial sphere and telling the world — we take our product division just as seriously as our research.
