Archive — US — techcrunch
DeepMind CEO Demis Hassabis says Google will eventually combine its Gemini andVeo AI models
DeepMind CEO Demis Hassabis announced Google's intent to merge its Gemini and Veo AI models to enhance understanding of the physical world, aiming for an omni-model AI that synthesizes various media types. The integration will leverage YouTube's vast video content for training, reflecting trends in AI towards universal digital assistants.
Key points
- Demis Hassabis discussed plans for merging Gemini and Veo AI models on a podcast.
- The goal is to create a universal digital assistant that assists users in real life.
- Gemini is designed as a multimodal foundation model capable of generating audio, images, and text.
- The merging of models is part of a broader trend in AI towards omni models synthesizing multiple media forms.
- Veo’s video data primarily comes from YouTube, allowing Google to enhance understanding of physics through video.
Why it matters
- The shift towards omni-models reflects a significant trend in AI development where models can handle diverse content.
- Historically, AI models have evolved from unidimensional functionalities to more complex integrations.
- As of 2025, companies like Amazon and OpenAI are also pursuing similar multidimensional AI models, indicating a competitive surge in the industry.
Google's plan to integrate Gemini and Veo models signifies a pivotal shift toward advanced omni-models in AI, linking diverse content for real-world applications, which is increasingly relevant in the evolving landscape of technology.