How can enterprises quickly adopt gen AI?
In this InTechnology video, Camille talks with co-host Stephanie Cope, Portfolio Development Manager at Intel Capital, and guest Rodrigo Liang, Co-Founder and CEO at SambaNova Systems. They get into SambaNova’s mission, how they help enterprises quickly scale AI, and the language aspect of LLMs.
Learn more about Intel Capital here.
SambaNova’s Mission
Rodrigo explains to Stephanie and Camille the founding and mission of SambaNova. He recounts starting the company in 2017 with two Stanford University professors, focusing on the entire stack—from models down to chips. Their goal was to build a superior computing platform that enables enterprises to run larger AI models trained on private data and perform inference at scale. This vision has now been realized. Rodrigo highlights that significant investments, particularly from Intel Capital, have been crucial in developing their chips, servers, compiler stack, models, and the trillion-parameter LLM, Samba-1. The company now offers reduced costs, improved chip performance, and simplified deployment for running models on fewer chips. Stephanie iterates on Intel Capital’s enthusiasm, noting their admiration for SambaNova’s comprehensive approach to creating a custom AI accelerator and offering it as a service to enterprises.
Then, Rodrigo delves into the details of the Samba-1 model, emphasizing its ability to keep pace with new model generations and integrate other open-source models. SambaNova’s strategy involves providing a complete hardware and software package, including the Samba-1 model. While training generative AI models can be extremely costly, SambaNova leverages the open-source community by pre-training Samba-1 on top-tier open-source models, offering a more cost-effective solution. Samba-1 integrates over 90 expert pre-trained models, resulting in enhanced performance, better response times, and quicker model improvements.
How Enterprises Can Quickly Scale AI
The discussion then shifts to implementing and scaling AI at the enterprise level with SambaNova. Rodrigo explains their focus on Global 2000 companies or any business with private data that can drive business transformation through gen AI models. Instead of investing in the necessary hardware, time, and personnel to train models themselves, enterprises can achieve similar results with Samba-1 in just a few days. Rodrigo notes a common challenge: enterprises want to leverage open-source models with their private data but retain ownership of that data. Samba-1 addresses this by allowing customers to fine-tune pre-trained, open-source models with their private data while maintaining ownership.
Rodrigo attributes the scalability of SambaNova’s solutions to the expert composition in Samba-1, which enables larger models to run on a smaller hardware footprint. Their hardware platform supports hundreds of models simultaneously within the same memory footprint and device, allowing for rapid model swapping. This capability, equivalent to what hundreds of GPUs would require, supports multiple users concurrently rather than one at a time. Additionally, Samba-1 integrates seamlessly into a customer’s secure environment, providing a fast, privately run ChatGPT-like generative AI experience.
Language and LLMs
Lastly, Rodrigo discusses the future of gen AI and LLMs in the context of language. While many LLMs today are trained primarily in English, he stresses that AI should not be limited to English speakers. He cites SambaNova’s work with languages like Hungarian, Japanese, and Thai for international clients. Rodrigo envisions a future where AI and commerce are driven by language, emphasizing the need for models to be segmented and regionalized for different languages and cultural contexts to achieve widespread adoption of gen AI.
Stephanie Cope, Portfolio Development Manager at Intel Capital
Stephanie joined Intel Capital in 2022, where she currently focuses on accelerating product development and market adoption for portfolio companies, including SambaNova. Prior to this role, she led Strategic Tech Innovation in Intel’s Health and Life Sciences division. Her extensive experience at Intel also includes positions as a Yield Engineering Manager and Yield Engineer. Before her tenure at Intel, Stephanie served as a Regional Manager and Application Physicist at Wyatt Technology and worked as a Research Associate at Arizona State University. She holds a Ph.D. in Physics from Arizona State University.
Rodrigo Liang, Co-Founder & CEO at SambaNova Systems
Rodrigo has been Co-Founder and CEO of SambaNova since 2017. Before founding SambaNova, he served as Senior Vice President at Oracle, where he led teams in designing microprocessors and ASICs for the Sun product line. He also held the position of Vice President at Sun Microsystems. Rodrigo holds a Master’s degree in Electrical Engineering from Stanford University.
Check it out. For more information, previous podcasts, and full versions, visit our homepage.
To read more about cybersecurity topics, visit our blog.
#generativeAI #genAI #SambaNova #IntelCapital
The views and opinions expressed are those of the guests and author and do not necessarily reflect the official policy or position of Intel Corporation.
—–
If you are interested in emerging threats, new technologies, or best tips and practices in cybersecurity, please follow the InTechnology podcast on your favorite podcast platforms: Apple Podcast and Spotify.
Follow our host Camille @morhardt.
Learn more about Intel Cybersecurity and Intel Compute Life Cycle (CLA).