OpenAI is COOKED

@phonggt_

141.3K followers · usually 36.5K views a post

Views

389.1K

Likes

16.1K

Comments

456

Shares

1.2K

SoundTikTok sound · original sound - phonggt_

Why it worked

The video leverages a controversial, high-stakes opinion about a major industry player (OpenAI vs. Meta) to drive engagement through debate and technical authority.

Summary

The creator discusses the release of Meta's open-source Llama 3.1 model and argues that it poses a significant threat to OpenAI. He highlights the capability of synthetic data generation as a key advantage of the new model.

Structure

1Hook: OpenAI is cooked
2Explanation of Llama 3.1 release
3Definition of synthetic data generation
4Comparison of OpenAI's terms vs Meta's open source approach
5Conclusion: Meta will win over OpenAI

On-screen text

OpenAI is COOKED Meet Llama 3.1 8B

Transcript

OpenAI is done so Meta released Llama 3.1 right? What's important is that they released a 405 billion parameter version Llama 3.1 models completely open source and I've been saying this for months. Smaller models are the future. And guess what? The new model will allow you to do this thing called synthetic data generation. What's synthetic data generation? Instead of having to scrape data from the internet yourself or having to harvest operational data from your users through using your app, you can just create it using an LLM. Open AI strictly does not allow this. They don't like it when you use their output to train competing models or making models commercially available to third parties. Don't bet against Meta when it comes to open source tech. Majority of webdev is React. PyTorch destroyed TensorFlow. Just a matter of time before OpenAI loses.