Demis Hassabis, CEO of Google’s DeepMind, reveals that Gemin aims to outperform OpenAI’s ChatGPT.
Take aways:
- Technology: Gemini will combine techniques from DeepMind’s AlphaGo with large language models like GPT-4. It aims to add new capabilities such as planning and problem-solving.
- Development & Costs: The system is still in development and could cost tens to hundreds of millions of dollars. OpenAI’s GPT-4 itself cost more than $100 million to develop.
- Competition: Gemini could be Google’s answer to the rising popularity of ChatGPT and other generative AI technologies. Google has already released its own chatbot, Bard.
- Reinforcement Learning: DeepMind plans to leverage its expertise in reinforcement learning to give Gemini novel capabilities.
- Safety & Ethics: Hassabis acknowledges the potential risks and ethical considerations in AI development. DeepMind has an “AI safety” group to explore these issues.
- Future Plans: Hassabis suggests that more research is needed to understand the risks and capabilities of more advanced AI models. He is open to giving academia early access to these frontier models.
