Started as a research organization by some of the most brilliant minds in tech, how did OpenAI become the tech giant we know today? If you’re living under a rock, OpenAI is the company behind some groundbreaking Artificial Intelligence technologies and tools like GPT-3, GPT-4 Turbo DALL-E, and more. 

OpenAI has witnessed the leadership of influential entrepreneurs like Sam Altman and Elon Musk and stands at a valuation of $29 Billion today. Today, OpenAI is a rage. But do you know the ups and downs that the company faced in the early days? Are you aware of the philosophy on which this startup was based? Do you know that Sam Altman, the CEO of the company, owns zero equity in the company?

We will find out how this humble scientific exploration went on to be funded by Microsoft and became one of the highest-valued tech startups of 2023. So, let’s dive right into it.

The Birth of an Idea

The journey of OpenAI began in 2015 when Sam Altman, President of the prestigious startup accelerator Y Combinator, met with tech visionary Elon Musk at the luxurious Rosewood Hotel in Silicon Valley. 

They talked about the potential and dangers of AGI over dinner. Their concern was similar – AGI in the wrong hands might be a total disaster. When two geniuses meet, what do you expect to happen?

Their solution to this problem was to create a benign AGI that would benefit all humanity so no one would be able to corrupt it for ill intent.

The "Free for All" Vision

Musk and Altman assembled a group of top researchers and scientists in AI and discussed the creation of a research lab. This institution aimed at advancing AGI against commercial goals. From this vision, OpenAI was conceived: a non-profit that would not have its activities influenced by profitability.

The Brilliant Team of OpenAI

Working closely with Musk and Altman were Greg Brockman (Open AI’s CTO and former CTO of Stripe) and Ilya Sutskever (Chief Scientist and Machine Learning expert), who played pivotal roles in steering the organization. 

Additional team members included Vicki Cheung, Trevor Blackwell, Durk Kingma, Andrej Karpathy, Pamela Vagata, John Schulman, and Wojciech Zaremba as scientists and researchers. Besides the core employees, the board of directors includes renowned individuals like Pieter Abbeel, Yoshusha Bengio, Alan Kaye, Sergey Levine, and Vishal Sikka. Sam and Elon led this association as co-chairs.

Last but not least, Amazon Web Services (AWS), Greg, Elon, Reid Hoffman, Sam, Peter Thiel, Jessica Livingston, and YC Research promised an informal contribution or donation of one billion US dollars for this project. However, only a small fraction of that was received in the early years.

Culture and Mission

In its early days, OpenAI exuded energy, excitement, and a strong sense of purpose. OpenAI had a collective narrative that was significantly influenced by Elon Musk, who played an important role. Despite the slim chances, he strongly emphasized the possible effects that AGI could cause.

While promoting innovation in the company’s operations, the informal culture was sometimes unclear of exactly where it intended to go. A Google researcher, Dario Amodei, said he was bewildered by OpenAI’s mission, a shared feeling among staff during that time. However, Amodei joined OpenAI with his sister Danalela, who followed him after two years. 

The Evolution of OpenAI's Mission

Initial Challenges

Altman, together with Greg Brockmann and Ilya Sutskever, faced a formidable challenge: formulating an appropriate line of action. However, they originally planned to release their research into the public domain so the wise crowds could work together for safety and goodness.

The Need for Resources

To realize their ambitious mission, they needed substantial resources. Nonetheless, their financial budget hit a snag. Altman recalls trying to raise $100 million in a room full of Silicon Valley investors, a staggering amount for a nonprofit. An unknown investment amount came from a pool of shareholders, under Bedrock Fund Management. Later, in 2016, Y-Combinator also initiated a seed investment in OpenAI. 

In March 2017, after 15 months of operation, OpenAI's leadership recognized the need for more focus. Brockman and a select group of core members began crafting an internal document to outline a path to AGI. 

However, this process unveiled a significant challenge. As the team analyzed trends in the field, they realized that maintaining nonprofit status was financially unsustainable. The computational resources required to achieve breakthrough results doubled every 3.4 months, necessitating a new organizational model that could rapidly accumulate capital. 

Elon Musk’s Exit

Sam Altman stated in an interview that Elon Musk thought that OpenAI was falling behind other competitive AI research companies. In February 2018, Elon cut ties with OpenAI, stating a “conflict of interest” as the primary reason for the exit, as he had his own AI research wing at his organization, Tesla. 

However, he didn’t leave the role of a donor to the organization but never contributed any funds either to it thereafter. 

Breakthroughs

Amidst of all these challenges, OpenAI had two exceptional breakthroughs. In 2018, OpenAI's researchers trained a robotic hand called Dactyl to manipulate objects with impressive dexterity using reinforcement learning. 

This breakthrough demonstrated the potential of AI in robotics and fine motor control tasks. Then, OpenAI developed an AI system, OpenAI Five, which achieved remarkable success in playing the popular video game Dota 2. 

Five competed and won against professional human players and showcased the capabilities of AI in complex, team-based gaming scenarios.

The Birth of OpenAI LP

Despite the breakthroughs and conflicts, OpenAI still needed money to operate further and skyrocket its growth on the promising technology and talent potential it had. Unbeknownst to the public and most employees, OpenAI released its charter in April 2018 to reflect this new reality. 

While reiterating its core values, the charter subtly shifted its language to emphasize the need for resources. In March 2019, OpenAI underwent a structural transformation. They created a subsidiary called OpenAI LP, which aimed to secure investments without compromising the nonprofit's mission. 

The organization shifted away from its strictly nonprofit status by establishing a "capped profit" division, allowing for a 100-fold limit on investor returns overseen by a board affiliated with a nonprofit entity. During this pivotal moment of OpenAI, Sam Altman became its CEO.
‍
‍

Khosla Ventures was the first VC to contribute to OpenAI as an investor. They injected $50 million into OpenAI during a funding round, resulting in a valuation of $1 billion for the startup. 

Unveiling ChatGPT-2

In 2016, OpenAI's researcher Alec Radford made a breakthrough with a language model, laying the foundation for ChatGPT. Sam Altman and team introduced the GPT-2 (Generative Pre-trained Transformer 2) model in 2019. 

GPT-2 is a powerful language model that generates coherent and contextually relevant text. It garnered significant attention for its ability to produce high-quality text and raised discussions about the responsible use of AI-generated content.

GPT-2 was trained on BookCorpus, which consisted of the data of 7000 books, which is very small as compared to the dataset on which GPT-3 and 4 are trained. It used to be referred to as the “deepfake for text.” While not as advanced as its predecessor, GPT-2 was a capable text generator. 

For example, if you feed it the opening line of a novel, it could recognize its tonality and context and produce accurate content in continuation with that line. Similarly, if you input the first few paragraphs of an article, it can input quotes and mentions from the article, along with some made-up “artificial” information.  

Enters Microsoft 

Shortly after, on the 22nd of July 2019, OpenAI announced a substantial billion-dollar investment from Microsoft, including cash and Azure credits.

Unsurprisingly, this shift raised concerns that OpenAI was diverging from its original mission. Some questioned how a 100-fold return limit would prevent undue concentration of power. The move also sparked internal unease among employees. 

To address these concerns, the leadership developed an FAQ as part of a set of confidential transition documents. They aimed to reassure people that OpenAI remained trustworthy despite the changes.

Striking the balance

OpenAI was initially celebrated for its transparency and commitment to sharing research. However, as the quest for resources and competitive pressure intensified, the lab's culture shifted. Some viewed this transition as a departure from its original ideals.

However, OpenAI took a highly unconventional path to strike the right balance and stay true to its vision. OpenAI LP had a workforce of approximately 100 individuals after the 2019 pivot. 

They were divided into three key areas: Capabilities (which involves pushing the boundaries of what AI systems can accomplish), Safety (ensuring these systems are in alignment with human values), and Policy (which oversees the appropriate governance of such systems). 

OpenAI Nonprofit started to serve as the governing body for OpenAI LP, manages educational initiatives like Scholars and Fellows, and hosts policy-related programs.  Lastly, nobody knows what goes on behind closed doors. However, initiative matters! This initiative shows the conscious approach that Sam Altman and the team have towards technology and its associated people. 

OpenAI: As we know today

The year 2020 marked a period of consolidation and self-discovery for OpenAI. As they delved deeper into the complexities of AGI, it became clear that their day-to-day work should focus on developing new AI technologies rather than commercial products. 

They were crafting the building blocks for something monumental, and they recognized that their structure should provide flexibility for long-term returns. Still, the ultimate goal was the creation of a safe AGI.

Finally, in 2021, OpenAI's research endeavors bore fruit. Their dedication to language models, such as DALL-E, Chat GPT 3, and 3.5, created virality on the internet and a great success story.  

OpenAI made headlines yet again in April 2023, securing a total investment of $300 million, with influential entities like K2 Global, Sound Ventures, Thrive Capital Management, Andreessen Horowitz, Tiger Global, Sequoia Capital Operations LLC, and AH Capital Management stepping in. 

This colossal investment pushed OpenAI's valuation to a staggering $29 billion. It was a testament to the belief in OpenAI's vision to make AGI safe and accessible. 

2023 was yet another turning point. Microsoft pledged a staggering $10 billion to support OpenAI. In 2023, OpenAI also took a giant leap forward with the launch of ChatGPT 4. 

Building upon the foundations of its predecessors, this cutting-edge AI model showcased remarkable improvements in factuality and trustworthiness, representing a bold stride toward safer and more reliable AI systems.

What’s next for OpenAI?

OpenAI is primarily working on GPT-4 Turbo with new features improved pricing, and DALL-E 3 API. Here are some upcoming OpenAI developments:

• GPT-4 Turbo: A more capable language model with a 128K context window, cost-effectiveness, and knowledge up to April 2023.
• Function Calling Updates: Enhanced function calling features for efficient communication with AI models.
• Improved Instruction Following: GPT-4 Turbo excels at precise instruction-following and supports JSON mode for accurate output.
• Assistants API: Empowering developers to create agent-like experiences within applications, simplifying interactions and facilitating persistent threads.
• New Modalities: GPT-4 Turbo accepting images as inputs, DALL·E 3, and Text-to-speech (TTS) features for advanced applications.
• Customization and Lower Prices: Fine-tuning and customization programs, coupled with reduced prices, enhanced accessibility.
• Copyright Shield: Support and legal protection to customers facing copyright infringement claims, promoting confidence in the platform's use.
• Jukebox: An AI model designed to generate music. It can compose original songs in a wide range of music genres and styles. 
• MuseNet: MuseNet can create original compositions, offering musicians and creators a powerful tool for generating music with minimal effort.
  ‍
  ‍

Conclusion: What made OpenAI win?

The journey of OpenAI is all about calculated risks, mindful innovation, rational optimism, and fierce competition. Silicon Valley has had enough experience to know that comfortable paychecks and ESOPs don’t always drive innovation. In fact, Sam Altman has zero equity in OpenAI. 

Along with talent, there also has to be a culture and an underlying philosophy that drives the mission. The organization started as a humble non-profit research project based on the principle of free innovation without the magnet of profitability. 

For the first three years, that’s how they operated. When they had sufficient research, groundwork, and vision in place, they knew they had to pivot to make the best use of the technology they were building and discovering. 

Consider the cryptocurrency market. The reason why a lot of companies in that sector tanked recently was because the early players were not ready to move on from the philosophy of anti-centralization. 

Perhaps in the future, we will witness some kind of centralization in cryptocurrency and blockchain-driven businesses, and they will thrive. However, OpenAI was flexible about its agenda and pivoted at the right time. 

It also realized the value of incentivizing talent with better compensations and embracing capitalism. Secondly, while OpenAI’s team was developing, parallely Google was also developing DeepMind. In 2017, OpenAI allocated $7.9 million, which accounted for a significant 25% of its operational expenditures, exclusively for cloud computing services. 

To put this in perspective, during the same year, DeepMind incurred a much higher total expenditure of $442 million. Yet OpenAI became the first movers in the AI world. However, it’s also important to note that both organizations had different approaches & strategies, and the AI community benefits from that. The strategy adopted by Sam Altman and OpenAI is “blitzscaling.” 
‍
‍

Now, there are Neuralink, DeepMind, Boston Dynamics, and Anthropic, to name a few companies in the AI market. Hence, the battle isn’t over, but in my opinion, what OpenAI did right to be at the forefront of the AI world so far is bringing together and retaining great talent, pivoting at the right time, being self-aware, and striking a balance between fast-paced innovation and conscious advancements. 

It will be fascinating to see what happens in the AI world next!