New Delhi: In the field of AI, Meta is making a big leap with the release of its latest open-weight AI model, Muse Glimmer, specifically targeting personal computers. The shift provides a potential avenue for Meta to acquire some of the leading AI companies, like OpenAI and Google.
On August 10, Meta debuted its 30-billion-parameter model, Muse Glimmer. In contrast to many high-powered AI systems which are cloud-based requiring many GPUs, Muse Glimmer is powered with only a single consumer-grade GPU and can use a PC or Mac installed locally.
Meta is betting on OpenAI
The biggest difference in Meta’s strategy is its focus on open-weight AI. Meta has shared the weights of the model with developers, enabling them to download, analyse, modify and develop apps based on it.
That might be a significant benefit for the developers and businesses who want to avoid the complete cloud-based AI relationship. Meta has also made Muse Glimmer available under the Apache 2.0 licence, and created tools and documentation to enable developers to create AI agents locally.
Why local AI could become important
The primary functions of Muse Glimmer are AI agent tasks. This means that it can be used in tasks like keeping track of schedules, organising files, launching software, and handling longer tasks with minimal human intervention.
When utilizing the AI locally, there is also a reduction of the need for all requests to be sent to a separate Data Center. This may be one way to enhance privacy, decrease delays and minimize operation costs of some applications. That’s a different approach for companies that primarily host their AI services as a cloud-based service.
Meta wants developers on its side
While Meta’s immediate focus is likely to be ChatGPT or Gemini, the company has the biggest opportunity to be the one that finally renders those other services redundant. Rather, the company could foster a vast developer community of its open models. With thousands of apps, tools and customisable AI systems which can be developed with its technology, Meta can have greater impact in the AI game. At the same time, Meta has confirmed that it will open up the heavyweights of its AI model (Muse Spark 1.2).
Can Meta really beat OpenAI and Google?
It is premature to declare any of the two, namely, Meta or OpenAI and Google, as the winner. Both have robust AI models, many users and decent cloud infrastructure. Google, for instance, is also training AIs that are tailored for certain devices and applications.
But Meta is taking another route. Rather than just simply racing to the largest and most powerful cloud AI model, it is attempting to bring AI to the mainstream as locally downloadable and accessible.
If the approach works, Meta might be able to press OpenAI and Google with increasing domination of the open, locally run AI market. Rather than the only company to win an AI battle, though, it is the one with the largest ecosystem built around its technology that will probably win.
