The Data Exchange with Ben Lorica Bridging the Hardware-Software Divide in AI
Aug 29, 2024
Jay Dawani, CEO of Lemurian Labs, dives into the challenges of bridging hardware and software in AI development. He discusses how model size influences performance and hurdles in achieving artificial general intelligence. The conversation highlights the critical need for seamless integration between training and inference, as well as the complexities of AI deployment. Dawani also explores the future of supercomputing in AI and the importance of optimizing data representation, showcasing innovative strategies to enhance computational capabilities.
AI Snips
Chapters
Transcript
Episode notes
Early Foundation Model Experience
- Jay Dawani, CEO of Lemurian Labs, worked on early foundation models, including a 2B parameter model in 2018.
- Training this model required 512 GPUs, prompting him to investigate compute scaling for larger models.
Hardware-Software Disconnect
- Dawani observed a disconnect between hardware, software, and AI companies' thinking regarding compute needs.
- This gap hindered realizing the potential of AI models, leading him to explore compute architectures.
Scaling Laws and AGI
- Scaling laws suggest increasing data and compute enhances model capabilities, driving trends towards larger models.
- AI teams assume following these laws is necessary for achieving Artificial General Intelligence (AGI).