E111 - How to build a LLM - Ariel Ekgren

The 111th episode of the AI After Work Podcast features Ariel Ekgren, a distinguished Research Scientist focused on developing Large Language Models (LLMs) for Sweden and the Nordics. Ekgren, who is both a Research Scientist and Tech Lead at AI Sweden, shares insights on the breakthroughs in deep learning and Natural Language Understanding. The episode delves into various topics, such as the impact of GPT decoder-only architecture, reasoning in GPT models, the Q* algorithm's progress towards AGI, and the creation and challenges of GPT-SW3, a specialized LLM for the Nordic region. Additionally, the discussion covers potential use cases for GPT-SW3, the benefits of multilingual versus region-specific models, future steps for GPT-SW3, the future of AI in Sweden, and speculations on whether AGI might lead to a dystopian or utopian future.

Follow us on youtube: https://www.youtube.com/@aiawpodcast