In This Article:
Meta Platforms (META, Financials) is assessing the potential use of the Chinese artificial intelligence model DeepSeek in its generative AI tools for advertisers, a company employee told The Information.
The action comes as Meta keeps investigating many artificial intelligence models to improve its advertising capacity and handle continuous issues with AI-generated content accuracy.
According to Meta's vice president of commercial artificial intelligence, Clara Shih, DeepSeek is under consideration for its "high performance." The corporation has not yet decided on a final choice on using the model. According to a Meta spokesman, Meta regularly assesses a broad spectrum of artificial intelligence models to ascertain their fit for enhancing corporate applications.
Citing that AI-generated content and pictures sometimes need major changes before being used in advertisements, advertisers have voiced worries about the quality and dependability of Meta's current AI capabilities driven by Llama. This has prompted Meta to investigate several models that can provide better content generating efficiency and accuracy.
Based on benchmark data posted on its website, DeepSeekwhich attracted major industry interest upon introducing its open-source AI model R1claims to surpass top AI models created by OpenAI, Meta, and Anthropic. If DeepSeek's assertions are accurate, analysts have observed that this would indicate a significant progress in AI-driven content production.
The Chinese hedge fund High-Flyer supports the model and shows how well huge language models may be created on a limited budget with little human oversight. DeepSeek's ascent has spurred debates in Silicon Valley, where worries about Chinese artificial intelligence companies posing a competitive threat are growing. DeepSeek's launch has already clearly affected market mood; Nvidia (NVDA, Financials) lost $593 billion in market capitalization among more general industry concerns about changing AI competitiveness.
This article first appeared on GuruFocus.