Last night OpenAI held their 'Spring Update' event, ostensibly to show off GPT-4o which will be available to free users, but industry watchers strongly suspect was to grab attention before the opening of the Google I/O developer conference which starts today.
If you've not actually tried OpenAI's systems like ChatGPT, you might not know that the company has restricted access to its best tech to paying users until this point, so wider free access is an interesting power play, and one which might be even more interesting when we come to another interesting rumor in a moment.
The focus of the announcement, GPT-4o, is a step forward from GPT-4, we were told, though, of course, it's difficult to know the realities of an AI model's features – what it definitely can and cannot do – given there are 'rough edges' to these things.
The big change with GPT-4o is that the system is able to reason using voice, text, and vision. That means, for example, it can detect tone when trying to translate (rather than transcribing to text and then attempting to translate the words).
In the demo, Mark Chen of Chat GPT received coaching for his breathing from a very lifelike voice which he could interrupt in a much more conversational way. There was no noticeable lag (though they were using a cable internet connection to the iPhone), and it was clearly picking up on (exaggerated) emotion.
The demo went on to use the phone's camera as the AI acted like a teacher, helping solve an equation with patience I'm not sure every teacher can boast, and recognising human emotion (after a little confusion between human and table).
Artificial intelligence, then, potentially, will depend a lot on phone cameras going forward. You can see the full demo here:
GPT-4o will be made immediately available to the company's API, which allows people to create their own apps that use AI (so it could be operating a call centre – or, indeed, an internet scam – near you).
Apparently, in part because of the risks, the features will be worked after discussions with governments. No details whatsoever were offered about this, however, and governments seem incompatible with the 'roll out over the next weeks' that was promised.
Oh, and the rumor? Bloomberg Technology has reported that Apple is close to a deal with OpenAI to put tech on Apple devices. There is a lot of discussion about AI being a selling point for phones, and Apple falling behind, so deal with a company like OpenAI (closely related, of course, to Microsoft) could help them catch up, at least in the short term.
What is the best camera phone? It might affect your AI!