Reinforcement Learning and Altman

Sam Altman at Bloomberg Summit in San Francisco

Front row with Sam Altman of OpenAI at the Bloomberg Technology Summit with host and executive producer Emily Chang!

Sam’s superpower is that he is a strategic thinker—he is able to look around the corner faster than anyone else in our technology industry. We all aspire to do this in Tech!

Mission—“I am excited to talk about what could happen in a few years and decades with AI technology.” This is how strategic leaders start and end their days. Strategic leadership means that you always think first about the big picture, go below the surface, examine statements with many perspectives, and look at the history of technology companies.

User Stories—What was surprising to Sam about the world tour he just did was learning how diverse users’ needs were and what they wanted ChatGPT to do. As a ChatGPT user who uses it far more than search engines, I use it for shaping my communications, teaching me about new technologies, and learning about experts I should connect to. It also has been helping me code an OpenAI-powered app on Python and React.js. These use cases are spread wide: I wanted to learn more.

Beneficial AI—Sam believes the long-term benefits of AI are better quality education, global medical care, and scientific progress, which will all improve our quality of life.

AI Regulation—Sam says that a global and coordinated response from AI companies should exist, but we should not regulate small startups. AI companies must have external quality audits just like any other industry.

Microsoft OpenAI Partnership—Sam said that “it was a crazy thing for us to jump into, and it’s going great.” Emily pressed him whether Google was concerned about competition, and Sam said that he would never write Google off. Mustafa Suleyman, Inflection AI‘s CEO, had something more controversial to say about Big Tech having built LLM-based chatbots but not getting it to market like OpenAI did.

Trust & Safety—Sam said that Reinforcement Learning, which GPT4 is built on, works “quite well to reduce bias. Technology has gone much further to align models to behave in certain ways.” Having met many Trust & Safety experts in recent months, I am skeptical about this.

Leadership—Sam has no equity in OpenAI and confirmed he has no financial incentives if OpenAI does well. “The concept of having enough money is hard to get across to others … I have enough money, I want an interesting life, and I want to contribute this technology to the world”. This is a remarkable leader.

Next up: Reid Hoffman and Mustafa Suleyman, and then Adam Selipsky, CEO of AWS.

Terrific production by Arena Choi and the Bloomberch Live Experiences team in San Francisco (still a great place for tech events!). Corporate event teams, connect with Arena to supercharge your events.

Reinforcement Learning: a Reading List

Raman Jha provided this reading list on Reinforcement Learning:

  1. OpenAI’s Spinning Up in Deep RL:
    Spinning Up in Deep RL is a fantastic resource for those looking to learn about deep reinforcement learning. It provides clear explanations, practical code examples, and a collection of algorithms implemented in popular deep learning frameworks like PyTorch.
    Website: Welcome to Spinning Up in Deep RL!
  2. David Silver’s Reinforcement Learning Course:
    David Silver, a researcher at DeepMind, has a comprehensive video lecture series on reinforcement learning. It covers both fundamental concepts and advanced topics in RL.
    YouTube Playlist: DeepMind x UCL | Introduction to Reinforcement Learning 2015
  3. Berkeley’s Deep Reinforcement Learning Course:
    UC Berkeley offers an online course on deep reinforcement learning that covers a wide range of topics, including policy gradients, actor-critic methods, and more.
    Course Website: CS 285
  4. Reinforcement Learning Specialization on Coursera (offered by the University of Alberta):
    This specialization is a series of courses that cover the foundations of reinforcement learning, as well as more advanced topics like value-based methods and policy-based methods.
    Course Link: Reinforcement Learning
  5. Reinforcement Learning with Python (book by Max Lapan):
    Max Lapan’s book provides a practical and hands-on introduction to reinforcement learning using Python. It covers both basic and advanced RL techniques.
    Book Link: https://www.packtpub.com/product/reinforcement-learning-with-python/9781789956055
  6. Sutton and Barto’s “Reinforcement Learning: An Introduction” (Book):
    This classic textbook by Richard S. Sutton and Andrew G. Barto is widely regarded as one of the foundational texts in the field. It’s an excellent resource for gaining a deep understanding of reinforcement learning.
    Book Link: Reinforcement Learning: An Introduction

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.