It seems that every iteration of ChatGPT and similar technologies is amazing, and the new “GPT-4o” certainly continues that trend. If you’re not familiar with it, here is a quick peek:

Along with just text, you can see that GPT-4o also can interpret video, images, and other info. Not only can it understand more about the world around it, but the overall interaction feels more lifelike than we’ve ever seen before.

As of today most of us still don’t have access to the video aspects of GPT-4o, but the audio alone is stunning. Here is another video that shows a bit more about that:

The ability for the AI to change the style of its voice like that is very compelling, and starts to get almost freaky good.

The voice model that we have access to today isn’t quite as powerful as shown in the video (that should be here very soon), but it’s still amazing. In playing with it a lot over the past few days, I’ve noticed two things:

1. The conversations that I have with it can be fantastic. I’ve talked to it about books and movies and the results are incredibly realistic and helpful.
2. While it has access to the real-time internet, it still struggles with some basic things. I tried to talk about yesterday’s Braves game, and it cited all kinds of incorrect information. As with previous models, it has no problem just creating fake “facts” if it’s unsure of the correct answer.

Where this will get more interesting in the coming months is when it can actually know me. Right now, GPT-4o can’t see my calendar or email or anything like that, so I can’t converse with it about my day. That’s undoubtedly going to change soon, possibly with some AI announcements from Apple at WWDC next month.

You can try GPT-4o today for free at ChatGPT.com, though your usage of it will be quite limited on a free account. Either way, this is a shocking improvement on an already amazing product, and will become even moreso as they release the other pieces of it (video, etc) in the coming weeks. Give it a shot.