GPT-4o is here and it's free. Here's why

OpenAI just launched GPT-4o. Ever wondered why they're giving it away for free? Here's what I think.

OpenAI earlier today launched a new, cutting-edge AI model.

GPT-4o now available

Touted as a step towards more natural-computer interaction, the GPT-4o model:

• Is faster than GPT-4.
• Listens, looks, and talks.
• Offers real-time reaction times.
• Matches GPT-4 Turbo in English, code.

In a demo, GPT-4o was shown holding conversations with OpenAI research leads and even reading the expression using the phone's camera to judge their emotions.

In a nutshell, it is another jaw-dropping release.

ChatGPT Plus (or minus)

GPT-4o is available to all, unlike GPT-4, which remain only accessible through the paid ChatGPT Plus service.

So what does ChatGPT Plus get you these days?

• Create Custom GPTs.
• 5x message limits for GPT-4o.
• Access to DALL-E 3, GPT-4 Turbo.

Not too shabby if you need these capabilities, though the need to subscribe is much lower now given how good GPT-4o is.

Why would OpenAI offer such a valuable model for free instead of charging for it? I have a theory.

Why is GPT-4o free?

While OpenAI could have attracted a lot more subscribers at US$20 per user, it opted to eschew it for a simple reason: It was never about the money.

Instead, it's about two things:

• AI market share

For OpenAI, gaining AI market share is its strategy for staying competitive, attracting investment, and shaping the future of generative AI.

And what better way than with a free GPT-4o, like it had with GPT-3.5, especially if it's cheaper to run than GPT-4. In fact, I suspect OpenAI only charged for GPT-4 due to its cost.

• Wealth of data

The other reason is the sheer volume of data OpenAI will get for being first to market with an easy-to-use multimodal AI optimised for conversations and vision.

Yes, you can opt not to have your data used to train future AI models. But do most users bother? And also... see my next point.

𝗣𝗦: Do you know where the setting is?

Is your data truly safe

Finally, it needs to be said. Can we really trust OpenAI to not use our data because we set a toggle on ChatGPT?

As I wrote previously, a NYT report fingered tech giants for using YouTube videos, copyrighted books, and other online content to train their AI models. OpenAI's Whisper was apparently created to pilfer YouTube videos - read about it here.

Of course, I don't know how OpenAI plans to use the data and interactions gleaned from GPT-4o.

But OpenAI doesn't have the best track record. And I do know that tech giants are running out of high-quality data for their models.

Is your chat data truly safe?