Meryem Arik: Women in Tech, Making LLM deployment efficient, LLMs for business | D.R.E.A.M
Podcast Link: • D.R.E.A.M (2024)
Meryem Arik, the co-founder and CEO of TitanML, discusses the importance of efficient inference in AI and the challenges of the current cost structure. She shares her background, childhood curiosity, and the influence of her entrepreneurial family. Meryem explains the transition from the corporate world to entrepreneurship and the evolution of generative AI models. She compares API-based and self-hosted models, highlighting the benefits of the self-hosted models provided by TitanML. Meryem also discusses the future of self-hosted language models and the need for radical change and fair distribution in the AI industry.
TIMESTAMPS:
00:00 Introduction and Purpose
01:29 Background and Mission of TitanML
03:01 Understanding Inference in AI
04:12 The Importance of Efficient Inference
05:19 Challenges of AI Cost Structure
06:29 Meryem's Background and Childhood Curiosity
08:48 Influence of Entrepreneurial Family
09:56 Transition from the Corporate World to Entrepreneurship
11:41 Meeting Co-Founders and Entry into AI
13:33 Evolution of Generative AI Models
15:23 Rapid Advancements in AI Technology
17:03 Concerns about Data Ownership and Control
18:40 Comparison of API-Based Models and Self-Hosted Models
19:36 Benefits of Self-Hosted Models
21:25 Infrastructure Provided by TitanML
23:18 Constraining Generation in Language Models
25:57 Future of Self-Hosted Language Models
27:43 Importance of Knowledge Bases and Retrieval
29:37 Vision for the Future and Policymaker Action
32:15 Need for Radical Change and Fair Distribution
35:09 Entrepreneurial Responsibility and Bigger Mission
36:25 Conclusion and Follow Meryem on LinkedIn