Unlocking Hidden Secrets of OpenAI

Table of Contents

1. Introduction
2. Surprising Fact: Open AI's Small Employee Size
3. Microsoft's Stake in Open AI
4. Changing Leadership at Open AI
5. Open AI's AI Bots Mastering Dota 2
6. Transition from Open Source to Closed Source
7. The Evolution of GPT Models
8. Open AI's Board Members
9. Elon Musk's Conflict with Open AI
10. High Salaries at Open AI
11. The High Training Costs of GPT Models
12. The Mystery of GPT4's Parameters
13. Open AI's Accomplishments in Robotics
14. Board Members Stepping Down and Lawsuits
15. Open AI's Corporate Structure
16. Open AI's KPI: Technical Progress
17. The End of Easy Scaling for AI Models
18. The Potential of Synthetic Data
19. The Departure of Elon Musk
20. Open AI's Projected Revenue

Introduction

Open AI is a rapidly growing company that has gained significant Attention in recent years. However, most people are still unaware of many intriguing facts about this company. In this article, we will explore 20 surprising facts about Open AI, shedding light on its size, ownership, leadership, achievements, challenges, and more. So, let's dive into the lesser-known aspects of Open AI.

1. Surprising Fact: Open AI's Small Employee Size

Despite its remarkable success, Open AI operates with only 375 employees. This is quite astonishing when You consider the company's accomplishments. Sam Altman, the CEO of Open AI, acknowledges this fact and emphasizes the exceptional talent density within the organization. Comparing Open AI's employee count to other tech giants like Google or Meta showcases the magnitude of their achievements.

2. Microsoft's Stake in Open AI

Microsoft holds a significant stake in Open AI, with an investment of $11 billion, giving them a 49% ownership share. As part of the agreement, Microsoft receives 75% of Open AI's profits until the $11 billion investment is repaid. This partnership not only benefits Microsoft but also grants Open AI access to Microsoft's supercomputers, providing them with a competitive edge.

3. Changing Leadership at Open AI

Greg Brockman, Open AI's former CTO, has been succeeded by Mira Murati, an AI engineer with a background in finance and experience at Tesla. Murati joined Open AI in 2018 and has now taken over as the Chief Technology Officer, overseeing the development of Open AI's cutting-edge AI model, GPT.

4. Open AI's AI Bots Mastering Dota 2

Several years ago, Open AI created a team of five AI bots that achieved mastery in the complex game Dota 2. These bots surpassed even the best human esports players in the game. Through self-play, these bots played the equivalent of 180 years' worth of games daily, showcasing a significant milestone in AI development.

5. Transition from Open Source to Closed Source

Contrary to what the name suggests, Open AI is not open-source. The company initially aimed to be open-source but shifted its focus to prioritizing safety. Open AI's decision to limit openness is primarily driven by the responsibility associated with deploying powerful AI models like GPT-4 without proper safeguards. Open AI's partnership with Microsoft, which sought exclusive rights to GPT-3, marked the end of its open-source nature.

6. The Evolution of GPT Models

The highly acclaimed GPT-3 model that powers the free version of ChatGPT was preceded by GPT-2 and GPT-1. Interestingly, GPT-3's predecessor, GPT-2, existed for almost two years before the creation of the highly praised GPT-3. The delay in releasing GPT-2 was due to its limited accessibility, as it was only accessible through an API, making it less user-friendly for the general public.

7. Open AI's Board Members

Open AI currently has seven board members, none of whom are democratically elected. The board includes individuals such as Shivan Zillis, Matthew Mokhari, Alejandro Cortufo, Jade Long, Kelly Sims, Helen Toner, and Atlas Benjalo. Notably, Shivan Zillis, one of the board members, is the mother of Elon Musk's twins, suggesting a connection between Elon Musk and Open AI.

8. Elon Musk's Conflict with Open AI

Elon Musk, a prominent figure associated with the founding of Open AI, cut off the company's access to Twitter data due to dissatisfaction with the arrangement. Musk believed that Open AI's payment of $2 million annually to Twitter for data access was insufficient. He publicly criticized the performance of ChatGPT and questioned the company's motives, referring to Open AI as a "maximum profit company."

9. High Salaries at Open AI

The salaries of top AI researchers at Open AI surpass those of many professional athletes. Even when Open AI operated as a non-profit organization, several top researchers earned over one million dollars annually. In 2016, the Chief Scientist, Ilya Satskiver, received a salary of $1.6 million. These high salaries illustrate the competitive remuneration offered by Open AI to attract and retain top talent.

10. The High Training Costs of GPT Models

The training costs associated with developing GPT-4 exceeded $100 million, a significant increase compared to the $12 million spent on training GPT-3. Such costs indicate the substantial investment required to develop and refine AI models. While it is unclear how the $100 million is allocated, some speculate that around $40 million is allocated for compute costs, with the remaining amount covering research salaries and costs associated with fine-tuning the model.

Continued in article...