Mark Zuckerberg Unveils New Product: What Can Meta's New AI Do?
Meta's limited rollout of its new models and claims of superiority in internal tests have drawn the attention of technology watchers and investors.
Meta Company unveiled two new artificial intelligence models named "Muse Wonder" and "Muse Wonder Full"; The models, which the company says are designed to perform complex tasks, program and act as intelligent assistants, are supposed to strengthen Meta's position in competition with OpenAI and Anthropic. These two new models are part of Meta's effort to close the gap with the AI industry leaders. Meta CEO Mark Zuckerberg has invested heavily in this area over the past year, recruiting leading researchers to develop more advanced models.
Meta says the new models can help users in areas such as finance, law and health. Among the intended applications of these models are the review of rental contracts, follow-up of insurance requests and software development. These systems are also equipped with active memory capabilities to retain important information from previous conversations for subsequent tasks. Another capability called "active focus" allows these models to try alternative solutions when faced with problems, correct errors, and strive to complete multi-step tasks. In this way, Meta's goal goes beyond answering users' questions and pursues a set of tasks with less human intervention.
The "Muse Wonder" model has been made available to developers through the Meta programming interface and the "Muse Code" programming tool. A more powerful version, "Muse Wonder Full", is also available to subscribers of the Muse Code tool. Meta has announced that the Wonder model will power the company's personal AI assistant in the future and will be available to subscribers first. This offering could create a new way to monetize Meta's heavy investments in artificial intelligence; Especially through paid subscriptions and offering services to businesses. However, performance claims of these models compared to competitors' products are mainly based on internal meta-tests, and their superiority in all applications has not been independently verified.
Changing the meta path in the development of artificial intelligence
The release of these models also indicates Meta's greater focus on commercial products; The path that is different from the previous approach of this company in the wide release of open source models. In this business model, access to the most advanced features is offered through paid services. At the same time, Meta is facing challenges such as the protection of personal information and the security of smart assistants that can access users' messages, calendars, and documents. The company says that in designing new models, it has taken into account the protection of private information and dealing with abuse; However, independent evaluation of their performance and safety is still important.
In 2023, Meta tried to establish its position in the generative artificial intelligence market with the release of the Llama family of language models. Initially, this company made it possible to build various tools based on its technology with the approach of publishing accessible models for researchers and developers. However, the performance of new models from companies such as OpenAI and Anthropic took the competition to a new level; A stage in which the ability to independently perform complex tasks and provide business services has become more important. Meta is also trying to gain more share of this market by increasing investment in computing infrastructure and attracting outstanding professionals.