On Tuesday, Meta held its first AI developer conference, Ramacon, at its headquarters in Menlo Park, California. The company has announced the launch of a meta AI chatbot app for consumers that competes with CHATGPT, as well as a developer API to access llama models in the cloud.
Both releases aim to expand the company’s adoption of the Open Lama AI model, but the goal may be secondary to breaking Openai, the true motivation of the meta. Meta’s AI ambitions promote a thriving open AI ecosystem that sticks to “closed” AI providers like Openai gated the models behind the service with broad strokes.
Meta’s AI chatbot app feels like the preemptive move for Openai’s rumoured social networks. It has a social feed that lets users share AI chats and provides personalized responses based on the user’s meta app activity.
When it comes to Llama APIs, it’s a challenge for Openai’s API business. The Llama API is designed to allow developers to build apps that connect to Llama models in the cloud using only one line of code. Eliminate the need to rely on third-party cloud providers to run the Llama model, allowing Meta to provide tools for AI developers in greater detail.
Meta, like many AI companies, recognizes Openai as a top rival. The court filing of the case against Meta reveals that he was previously obsessed with defeating Openai’s GPT-4, which was once a cutting-edge model. Filling up unique AI model providers like OpenAI was at the heart of Meta’s AI strategy. In a July 2024 letter, Meta CEO Mark Zuckerberg tried to contrast it with companies like Meta and Openai, writing, “Selling access to AI models is not a business model (Meta’s)”.
AI researchers who spoke with TechCrunch before Llamacon wanted Meta to release a competitive AI inference model like Openai’s O3-Mini. The company didn’t end up doing that. However, for Meta, it’s not necessarily about winning AI races.
In an on-stage conversation with Databricks CEO Ali Ghodsi at Llamacon, Zuckerberg said he views AI labs that make models that include Deepseek and Alibaba’s Qwen openly available, as an ally in the fight against closed model providers.
“Part of the value when it comes to open source is that you can mix and match. So if another model like Deepseek is good, or if Qwen is good at something, as a developer, you have the ability to take the best part of intelligence from the various models and generate exactly what you need,” Zuckerberg said. “This is part of the way that open source basically passes the quality of all closed sources (models)… (i)t feels like an unstoppable force.”
In addition to stopping Openai’s growth, it is also possible that Meta is trying to push open models to meet the regulatory carve-out. The EU AI Act gives companies special privileges to distribute “free and open source” AI systems. Meta often argues that the llama model is “open source” despite differences of opinion about whether it meets the required criteria.
Regardless of the reason, Meta appears happy to start launching AI. This is to strengthen the open model ecosystem and limit Openai’s growth.