My Reflections on Open AI's Spring Update
My Reflections on Open AI’s Spring Update.
Anybody that says, Open AI has lost any luster did not see the excitement online for Today’s announcement.
I like the fact that Sam Altman didn’t present and let some of the other Open AI higher-ups deliver the Spring Update presentation.
The big announcement is a model called GPT 4o, the O stands for Omni. It’s essentially Scarlett Johannsen from the movie Her. It is a more natural-sounding AI that can see, hear, and even view images. Much of the AI community has complained about ChatGPT’s voice feature being slow now that latency has been significantly improved with an average request speed of 320 milliseconds.
Many thought this event would be an announcement of a search engine from Open AI but that turned out to be a false rumor and possibly planted information. So they can catch who has been leaking information to OpenAI leakers Flowers from the Future and the notorious Jimmy Apples on Twitter.
The fact that this model’s API is 2x faster and 50% cheaper is an instant value proposition for any developer within the AI developer ecosystem. I hope Google is seeing this.
The really impressive thing about the presentation’s demo of GPT 4o was the audio portion when the model could understanding the breathing of the presenter. That was pretty unreal and lets you know it's not just understanding words and phrases but any type of audio.
I couldn’t help but think about Vinod Khosla’s statement about how AI will allow every student to have a tutor. This is very true with GPT 4o because you can have it help you solve equations, and geography problems and tell you how to get to answers which is half the learning process.
An underrated thing about the GPT 4o model is that it can make 3D models.
According to Nick Dobos, the creator of the popular GPT, Grimoire. Voice has been switched to being a secondary system, to being actually integrated into the model.
Real-time translation with the GPT 4o model was so impactful that Duolingo’s stock dropped 5 percent.
Greg Brockman, President of Open AI, letting the two AIs talk to each other and even sing a song together with alternating lines, makes me hope that AI doesn’t become some Skynet sentient being and makes me sing a song about the robot apocalypse and alternate lines with my friend.
A lot of people talked about how this announcement killed a lot of AI companies in the process but I think the opposite it will make those companies better and want to pivot to the Open AI API even more. Remember, “developers, developers, developers”.
The overall human “vibe” of the 4o model is what science fiction movies are made of.