Lightning Interview “Large Language Models: Past, Present and Future”
Creator of Llama 2, one of the world’s best-known Generative AI projects, Thomas Scialom, PhD, has been at the center of cutting-edge AI development. Currently, he is contributing to the development of Artificial General Intelligence (AGI). Thomas is also a well-known and respected lecturer on AI.
Topics
1– Tell us about your background and work as a researcher at Meta AI
2– Can you sum up the history of LLMs and some of the big breakthroughs?
3– What are some of the major lessons we can draw from the brief history of LLMs and GenAI
4– Can you explain the Chinchilla scaling laws in more detail and the future implications of this scaling law for the development of even larger LLMs
5– Describe how you and the team built Llama2 and how long did it take.
6– Can you elaborate on some of the biggest challenges you've encountered in developing large-scale AI projects, in particular, Llama 2 and Code Llama
7– What safety and ethical considerations went into developing thing modes and what methods were used?
8– What impact has the open sourcing of LLaMA 2 and its availability for free commercial use had on the LLM and AI ecosystem?
9– Can you tell us more about Code Llama which is a family of LLMs specifically designed for generating code and built upon LLaMA 2,
10– What do you think will happen next in LLMs over the next 12 to 24 months?
11– What are the key milestones we need to achieve to reach AGI?
12– How soon will we see the integration of LLMs and GenAI in everyday consumer technology?
13– Can you tell us about the startup ecosystem in France? There seems to be a lot of talented researchers and startup emerging in the last few years.
14– Where can people follow your work?
Useful Links:
Get in touch with Thomas via / thomasscialomopens or / tscialom
Here is the link to the model for Interview’s session - https://ai.meta.com/resources/models-...
Toolformer: Language Models Can Teach Themselves to Use Tools. Link to the paper you may find here - https://arxiv.org/pdf/2302.04761.pdf