Meta Aims to Build an Open-Source ChatGPT Contender

The artificial intelligence race has been reignited again!

This time around, Meta (Facebook/Instagram's parent company) gets the attention.

In the past, we have seen the likes of Mozilla betting big on developing an open-source AI ecosystem and the likes of Dolly and HuggingChat acting as open-source alternatives to ChatGPT.

However, with Meta's new undertaking, we might have something even more significant. Allow me to take you through it.

What's Happening: According to a recent report by The Wall Street Journal, Meta intends to take the fight to OpenAI's ChatGPT by building a new open-source AI system, expected to be as capable as GPT-4.

Interestingly, Meta is no stranger to developing an AI system; they already have the Llama source-available large language model co-developed with Microsoft, which recently received a new major release.

But this upcoming model is dubbed several times more powerful than that! 🤯

Moreover, an open-source approach to a potential ChatGPT replacement sounds exciting!

Other companies are expected to benefit from the system, allowing them to build services to produce sophisticated text, analysis, and other types of output.

Suggested Read 📖

When to Expect: The work has already begun, with Meta readying up the data centers necessary to pull off this job.

They have opted for NVIDIA H100 data-center GPUs, signaling a move away from using Microsoft's Azure, and focusing on training this new model on their infrastructure.

It is speculated that Meta will begin training the new model in early 2024, with Mark Zuckerberg leading the dedicated team to make it all happen.

Even though Meta has been late to the generative AI game, this new open-source AI system should put them right back into the mix, especially when Google is also launching a powerful competitor to GPT-4 called 'Gemini'.

💬 What are your thoughts on Meta's initiative for an open-source ChatGPT competitor?